X-Git-Url: http://www.privoxy.org/gitweb/?p=privoxy.git;a=blobdiff_plain;f=parsers.c;h=5f8c68616c649b68893f35f5c7818575496acfa0;hp=176fe5625b44d4aa0630bd321629ccb6c6a3d8cd;hb=9f662300f7cae2ffb1656f97a4dca0391f292d0f;hpb=23b99e99ad07aefdc868f16c970c22f73b7f50cc diff --git a/parsers.c b/parsers.c index 176fe562..5f8c6861 100644 --- a/parsers.c +++ b/parsers.c @@ -1,12 +1,11 @@ -const char parsers_rcs[] = "$Id: parsers.c,v 1.312 2017/06/08 13:13:26 fabiankeil Exp $"; /********************************************************************* * * File : $Source: /cvsroot/ijbswa/current/parsers.c,v $ * * Purpose : Declares functions to parse/crunch headers and pages. * - * Copyright : Written by and Copyright (C) 2001-2016 the - * Privoxy team. http://www.privoxy.org/ + * Copyright : Written by and Copyright (C) 2001-2020 the + * Privoxy team. https://www.privoxy.org/ * * Based on the Internet Junkbuster originally written * by and Copyright (C) 1997 Anonymous Coders and @@ -65,6 +64,9 @@ const char parsers_rcs[] = "$Id: parsers.c,v 1.312 2017/06/08 13:13:26 fabiankei #define GZIP_FLAG_COMMENT 0x10 #define GZIP_FLAG_RESERVED_BITS 0xe0 #endif +#ifdef FEATURE_BROTLI +#include +#endif #if !defined(_WIN32) && !defined(__OS2__) #include @@ -90,8 +92,6 @@ const char parsers_rcs[] = "$Id: parsers.c,v 1.312 2017/06/08 13:13:26 fabiankei #include "strptime.h" #endif -const char parsers_h_rcs[] = PARSERS_H_VERSION; - static char *get_header_line(struct iob *iob); static jb_err scan_headers(struct client_state *csp); static jb_err header_tagger(struct client_state *csp, char *header); @@ -252,13 +252,14 @@ static const add_header_func_ptr add_server_headers[] = { /********************************************************************* * - * Function : flush_socket + * Function : flush_iob * * Description : Write any pending "buffered" content. * * Parameters : * 1 : fd = file descriptor of the socket to read * 2 : iob = The I/O buffer to flush, usually csp->iob. + * 3 : delay = Number of milliseconds to delay the writes * * Returns : On success, the number of bytes written are returned (zero * indicates nothing was written). On error, -1 is returned, @@ -268,7 +269,7 @@ static const add_header_func_ptr add_server_headers[] = { * file, the results are not portable. * *********************************************************************/ -long flush_socket(jb_socket fd, struct iob *iob) +long flush_iob(jb_socket fd, struct iob *iob, unsigned int delay) { long len = iob->eod - iob->cur; @@ -277,7 +278,7 @@ long flush_socket(jb_socket fd, struct iob *iob) return(0); } - if (write_socket(fd, iob->cur, (size_t)len)) + if (write_socket_delayed(fd, iob->cur, (size_t)len, delay)) { return(-1); } @@ -392,6 +393,87 @@ void clear_iob(struct iob *iob) #ifdef FEATURE_ZLIB +#ifdef FEATURE_BROTLI +/********************************************************************* + * + * Function : decompress_iob_with_brotli + * + * Description : Decompress buffered page using Brotli. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * + * Returns : JB_ERR_OK on success, + * JB_ERR_MEMORY if out-of-memory limit reached, and + * JB_ERR_COMPRESS if error decompressing buffer. + * + *********************************************************************/ +static jb_err decompress_iob_with_brotli(struct client_state *csp) +{ + BrotliDecoderResult result; + char *decoded_buffer; + size_t decoded_size; + size_t decoded_buffer_size; + size_t encoded_size; + enum { MAX_COMPRESSION_FACTOR = 15 }; + + encoded_size = (size_t)(csp->iob->eod - csp->iob->cur); + /* + * The BrotliDecoderDecompress() api is a bit unfortunate + * and requires the caller to reserve enough memory for + * the decompressed content. Hopefully reserving + * MAX_COMPRESSION_FACTOR times the original size is + * sufficient. If not, BrotliDecoderDecompress() will fail. + */ + decoded_buffer_size = encoded_size * MAX_COMPRESSION_FACTOR; + + if (decoded_buffer_size > csp->config->buffer_limit) + { + log_error(LOG_LEVEL_ERROR, + "Buffer limit reached before decompressing iob with Brotli"); + return JB_ERR_MEMORY; + } + + decoded_buffer = malloc(decoded_buffer_size); + if (decoded_buffer == NULL) + { + log_error(LOG_LEVEL_ERROR, + "Failed to allocate %d bytes for Brotli decompression", + decoded_buffer_size); + return JB_ERR_MEMORY; + } + + decoded_size = decoded_buffer_size; + result = BrotliDecoderDecompress(encoded_size, + (const uint8_t *)csp->iob->cur, &decoded_size, + (uint8_t *)decoded_buffer); + if (result == BROTLI_DECODER_RESULT_SUCCESS) + { + /* + * Update the iob, since the decompression was successful. + */ + freez(csp->iob->buf); + csp->iob->buf = decoded_buffer; + csp->iob->cur = csp->iob->buf; + csp->iob->eod = csp->iob->cur + decoded_size; + csp->iob->size = decoded_buffer_size; + + log_error(LOG_LEVEL_RE_FILTER, + "Decompression successful. Old size: %d, new size: %d.", + encoded_size, decoded_size); + + return JB_ERR_OK; + } + else + { + log_error(LOG_LEVEL_ERROR, "Failed to decompress buffer with Brotli"); + freez(decoded_buffer); + + return JB_ERR_COMPRESS; + } +} +#endif + /********************************************************************* * * Function : decompress_iob @@ -447,6 +529,13 @@ jb_err decompress_iob(struct client_state *csp) return JB_ERR_COMPRESS; } +#ifdef FEATURE_BROTLI + if (csp->content_type & CT_BROTLI) + { + return decompress_iob_with_brotli(csp); + } +#endif + if (csp->content_type & CT_GZIP) { /* @@ -635,6 +724,7 @@ jb_err decompress_iob(struct client_state *csp) if (bufsize >= csp->config->buffer_limit) { log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob"); + freez(buf); return JB_ERR_MEMORY; } @@ -1187,6 +1277,65 @@ jb_err sed(struct client_state *csp, int filter_server_headers) } +#ifdef FEATURE_HTTPS_INSPECTION +/********************************************************************* + * + * Function : sed_https + * + * Description : add, delete or modify lines in the HTTPS client + * header streams. Wrapper around sed(). + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * + * Returns : JB_ERR_OK in case off success, or + * JB_ERR_MEMORY on some out-of-memory errors, or + * JB_ERR_PARSE in case of fatal parse errors. + * + *********************************************************************/ +jb_err sed_https(struct client_state *csp) +{ + jb_err err; + struct list headers; + + /* + * Temporarily replace csp->headers with csp->https_headers + * to trick sed() into filtering the https headers. + */ + headers.first = csp->headers->first; + headers.last = csp->headers->last; + csp->headers->first = csp->https_headers->first; + csp->headers->last = csp->https_headers->last; + + /* + * Start with fresh tags. Already existing tags may + * be set again. This is necessary to overrule + * URL-based patterns. + */ + destroy_list(csp->tags); + + /* + * We want client header filters and taggers + * so temporarily remove the flag. + */ + csp->flags &= ~CSP_FLAG_CLIENT_HEADER_PARSING_DONE; + err = sed(csp, FILTER_CLIENT_HEADERS); + csp->flags |= CSP_FLAG_CLIENT_HEADER_PARSING_DONE; + + /* + * Update the last header which may have changed + * due to header additions, + */ + csp->https_headers->last = csp->headers->last; + + csp->headers->first = headers.first; + csp->headers->last = headers.last; + + return err; +} +#endif /* def FEATURE_HTTPS_INSPECTION */ + + /********************************************************************* * * Function : update_server_headers @@ -1900,7 +2049,7 @@ static jb_err client_connection(struct client_state *csp, char **header) if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING) && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)) { - if (!strcmpic(csp->http->ver, "HTTP/1.1")) + if (!strcmpic(csp->http->version, "HTTP/1.1")) { log_error(LOG_LEVEL_HEADER, "Removing \'%s\' to imply keep-alive.", *header); @@ -2351,6 +2500,15 @@ static jb_err server_content_encoding(struct client_state *csp, char **header) /* Mark for zlib decompression */ csp->content_type |= CT_DEFLATE; } + else if (strstr(*header, "br")) + { +#ifdef FEATURE_BROTLI + /* Mark for Brotli decompression */ + csp->content_type |= CT_BROTLI; +#else + csp->content_type |= CT_TABOO; +#endif + } else if (strstr(*header, "compress")) { /* @@ -2401,7 +2559,7 @@ static jb_err server_content_encoding(struct client_state *csp, char **header) * * Description : Remove the Content-Encoding header if the * decompression was successful and the content - * has been modifed. + * has been modified. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) @@ -2417,7 +2575,12 @@ static jb_err server_content_encoding(struct client_state *csp, char **header) static jb_err server_adjust_content_encoding(struct client_state *csp, char **header) { if ((csp->flags & CSP_FLAG_MODIFIED) - && (csp->content_type & (CT_GZIP | CT_DEFLATE))) + && ((csp->content_type & (CT_GZIP | CT_DEFLATE)) +#ifdef FEATURE_BROTLI + || (csp->content_type & CT_BROTLI) +#endif + ) + ) { /* * We successfully decompressed the content, @@ -2690,9 +2853,8 @@ static jb_err server_last_modified(struct client_state *csp, char **header) time_t now; struct tm *timeptr = NULL; long int rtime; -#ifdef HAVE_GMTIME_R struct tm gmt; -#endif + now = time(NULL); rtime = (long int)difftime(now, last_modified); if (rtime) @@ -2711,15 +2873,7 @@ static jb_err server_last_modified(struct client_state *csp, char **header) rtime *= -1; } last_modified += rtime; -#ifdef HAVE_GMTIME_R - timeptr = gmtime_r(&last_modified, &gmt); -#elif defined(MUTEX_LOCKS_AVAILABLE) - privoxy_mutex_lock(&gmtime_mutex); - timeptr = gmtime(&last_modified); - privoxy_mutex_unlock(&gmtime_mutex); -#else - timeptr = gmtime(&last_modified); -#endif + timeptr = privoxy_gmtime_r(&last_modified, &gmt); if ((NULL == timeptr) || !strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr)) { @@ -2729,7 +2883,6 @@ static jb_err server_last_modified(struct client_state *csp, char **header) freez(*header); return JB_ERR_OK; } - freez(*header); *header = strdup("Last-Modified: "); string_append(header, newheader); @@ -3355,9 +3508,7 @@ static jb_err client_host(struct client_state *csp, char **header) static jb_err client_if_modified_since(struct client_state *csp, char **header) { char newheader[50]; -#ifdef HAVE_GMTIME_R struct tm gmt; -#endif struct tm *timeptr = NULL; time_t tm = 0; const char *newval; @@ -3415,15 +3566,7 @@ static jb_err client_if_modified_since(struct client_state *csp, char **header) *header); } tm += rtime * (negative_range ? -1 : 1); -#ifdef HAVE_GMTIME_R - timeptr = gmtime_r(&tm, &gmt); -#elif defined(MUTEX_LOCKS_AVAILABLE) - privoxy_mutex_lock(&gmtime_mutex); - timeptr = gmtime(&tm); - privoxy_mutex_unlock(&gmtime_mutex); -#else - timeptr = gmtime(&tm); -#endif + timeptr = privoxy_gmtime_r(&tm, &gmt); if ((NULL == timeptr) || !strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr)) { @@ -3433,7 +3576,6 @@ static jb_err client_if_modified_since(struct client_state *csp, char **header) freez(*header); return JB_ERR_OK; } - freez(*header); *header = strdup("If-Modified-Since: "); string_append(header, newheader); @@ -3841,7 +3983,7 @@ static jb_err client_connection_header_adder(struct client_state *csp) #ifdef FEATURE_CONNECTION_KEEP_ALIVE if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE) && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED) - && !strcmpic(csp->http->ver, "HTTP/1.1")) + && !strcmpic(csp->http->version, "HTTP/1.1")) { csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE; return JB_ERR_OK; @@ -3974,18 +4116,9 @@ static void add_cookie_expiry_date(char **cookie, time_t lifetime) char tmp[50]; struct tm *timeptr = NULL; time_t expiry_date = time(NULL) + lifetime; -#ifdef HAVE_GMTIME_R struct tm gmt; - timeptr = gmtime_r(&expiry_date, &gmt); -#elif defined(MUTEX_LOCKS_AVAILABLE) - privoxy_mutex_lock(&gmtime_mutex); - timeptr = gmtime(&expiry_date); - privoxy_mutex_unlock(&gmtime_mutex); -#else - timeptr = gmtime(&expiry_date); -#endif - + timeptr = privoxy_gmtime_r(&expiry_date, &gmt); if (NULL == timeptr) { log_error(LOG_LEVEL_FATAL, @@ -4328,9 +4461,10 @@ static jb_err parse_header_time(const char *header_time, time_t *result) { char recreated_date[100]; struct tm *tm; + struct tm storage; time_t result2; - tm = gmtime(result); + tm = privoxy_gmtime_r(result, &storage); if (!strftime(recreated_date, sizeof(recreated_date), time_formats[i], tm)) { @@ -4431,8 +4565,6 @@ jb_err get_destination_from_headers(const struct list *headers, struct http_requ char *p; char *host; - assert(!http->ssl); - host = get_header_value(headers, "Host:"); if (NULL == host) @@ -4485,7 +4617,88 @@ jb_err get_destination_from_headers(const struct list *headers, struct http_requ string_append(&http->cmd, " "); string_append(&http->cmd, http->url); string_append(&http->cmd, " "); - string_append(&http->cmd, http->ver); + string_append(&http->cmd, http->version); + if (http->cmd == NULL) + { + return JB_ERR_MEMORY; + } + + return JB_ERR_OK; + +} + + +#ifdef FEATURE_HTTPS_INSPECTION +/********************************************************************* + * + * Function : get_destination_from_https_headers + * + * Description : Parse the previously encrypted "Host:" header to + * get the request's destination. + * + * Parameters : + * 1 : headers = List of headers (one of them hopefully being + * the "Host:" header) + * 2 : http = storage for the result (host, port and hostport). + * + * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems, + * JB_ERR_PARSE if the host header couldn't be found, + * JB_ERR_OK otherwise. + * + *********************************************************************/ +jb_err get_destination_from_https_headers(const struct list *headers, struct http_request *http) +{ + char *q; + char *p; + char *host; + + host = get_header_value(headers, "Host:"); + + if (NULL == host) + { + log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found."); + return JB_ERR_PARSE; + } + + p = strdup_or_die(host); + chomp(p); + q = strdup_or_die(p); + + freez(http->hostport); + http->hostport = p; + freez(http->host); + http->host = q; + q = strchr(http->host, ':'); + if (q != NULL) + { + /* Terminate hostname and evaluate port string */ + *q++ = '\0'; + http->port = atoi(q); + } + else + { + http->port = 443; + } + + /* Rebuild request URL */ + freez(http->url); + http->url = strdup_or_die(http->path); + + log_error(LOG_LEVEL_HEADER, + "Destination extracted from \"Host\" header. New request URL: %s", + http->url); + + /* + * Regenerate request line in "proxy format" + * to make rewrites more convenient. + */ + assert(http->cmd != NULL); + freez(http->cmd); + http->cmd = strdup_or_die(http->gpc); + string_append(&http->cmd, " "); + string_append(&http->cmd, http->url); + string_append(&http->cmd, " "); + string_append(&http->cmd, http->version); if (http->cmd == NULL) { return JB_ERR_MEMORY; @@ -4494,6 +4707,7 @@ jb_err get_destination_from_headers(const struct list *headers, struct http_requ return JB_ERR_OK; } +#endif /* def FEATURE_HTTPS_INSPECTION */ /*********************************************************************