X-Git-Url: http://www.privoxy.org/gitweb/?p=privoxy.git;a=blobdiff_plain;f=filters.c;h=36eccf41e0e2690ca1657688dfc674c4c0acc898;hp=1a1da409823f7a6ed0df34740e5f3da4222b71b4;hb=7bfb6887ab00db9e025508bff29c122f05a63ace;hpb=656cce68348695162120584f5c12ceeed16b2792 diff --git a/filters.c b/filters.c index 1a1da409..36eccf41 100644 --- a/filters.c +++ b/filters.c @@ -1,4 +1,4 @@ -const char filters_rcs[] = "$Id: filters.c,v 1.63 2006/08/31 10:11:28 fabiankeil Exp $"; +const char filters_rcs[] = "$Id: filters.c,v 1.76 2007/01/01 19:36:37 fabiankeil Exp $"; /********************************************************************* * * File : $Source: /cvsroot/ijbswa/current/filters.c,v $ @@ -10,9 +10,10 @@ const char filters_rcs[] = "$Id: filters.c,v 1.63 2006/08/31 10:11:28 fabiankeil * `filter_popups', `forward_url', 'redirect_url', * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose', * `ijb_send_banner', `trust_url', `gif_deanimate_response', - * `jpeg_inspect_response' + * `jpeg_inspect_response', `execute_single_pcrs_command', + * `rewrite_url', `get_last_url' * - * Copyright : Written by and Copyright (C) 2001, 2004 the SourceForge + * Copyright : Written by and Copyright (C) 2001, 2004-2006 the SourceForge * Privoxy team. http://www.privoxy.org/ * * Based on the Internet Junkbuster originally written @@ -39,6 +40,58 @@ const char filters_rcs[] = "$Id: filters.c,v 1.63 2006/08/31 10:11:28 fabiankeil * * Revisions : * $Log: filters.c,v $ + * Revision 1.76 2007/01/01 19:36:37 fabiankeil + * Integrate a modified version of Wil Mahan's + * zlib patch (PR #895531). + * + * Revision 1.75 2006/12/29 18:30:46 fabiankeil + * Fixed gcc43 conversion warnings, + * changed sprintf calls to snprintf. + * + * Revision 1.74 2006/12/24 17:37:38 fabiankeil + * Adjust comment in pcrs_filter_response() + * to recent pcrs changes. Hohoho. + * + * Revision 1.73 2006/12/23 16:01:02 fabiankeil + * Don't crash if pcre returns an error code + * that pcrs didn't expect. Fixes BR 1621173. + * + * Revision 1.72 2006/12/22 18:52:53 fabiankeil + * Modified is_untrusted_url to complain in case of + * write errors and to give a reason when adding new + * entries to the trustfile. Closes FR 1097611. + * + * Revision 1.71 2006/12/22 14:24:52 fabiankeil + * Skip empty filter files in pcrs_filter_response, + * but don't ignore the ones that come afterwards. + * Fixes parts of BR 1619208. + * + * Revision 1.70 2006/12/09 13:33:15 fabiankeil + * Added some sanity checks for get_last_url(). + * Fixed possible segfault caused by my last commit. + * + * Revision 1.69 2006/12/08 12:39:13 fabiankeil + * Let get_last_url() catch https URLs as well. + * + * Revision 1.68 2006/12/05 14:45:48 fabiankeil + * Make sure get_last_url() behaves like advertised + * and fast-redirects{} can be combined with redirect{}. + * + * Revision 1.67 2006/11/28 15:19:43 fabiankeil + * Implemented +redirect{s@foo@bar@} to generate + * a redirect based on a rewritten version of the + * original URL. + * + * Revision 1.66 2006/09/23 13:26:38 roro + * Replace TABs by spaces in source code. + * + * Revision 1.65 2006/09/21 12:54:43 fabiankeil + * Fix +redirect{}. Didn't work with -fast-redirects. + * + * Revision 1.64 2006/08/31 10:55:49 fabiankeil + * Block requests for untrusted URLs with status + * code 403 instead of 200. + * * Revision 1.63 2006/08/31 10:11:28 fabiankeil * Don't free p which is still in use and will be later * freed by free_map(). Don't claim the referrer is unknown @@ -765,7 +818,7 @@ struct http_response *block_url(struct client_state *csp) if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0) && is_imageurl(csp)) { - char *p; + char *p; /* determine HOW images should be blocked */ p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER]; @@ -889,7 +942,6 @@ struct http_response *block_url(struct client_state *csp) return cgi_error_memory(); } } - } else #endif /* def FEATURE_IMAGE_BLOCKING */ @@ -1044,7 +1096,7 @@ struct http_response *trust_url(struct client_state *csp) p = strdup(""); for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++) { - sprintf(buf, "
  • %s
  • \n", t->spec); + snprintf(buf, sizeof(buf), "
  • %s
  • \n", t->spec); string_append(&p, buf); } err = map(exports, "trusted-referrers", 1, p, 0); @@ -1066,7 +1118,7 @@ struct http_response *trust_url(struct client_state *csp) p = strdup(""); for (l = csp->config->trust_info->first; l ; l = l->next) { - sprintf(buf, "
  • %s
    \n",l->str, l->str); + snprintf(buf, sizeof(buf), "
  • %s
    \n", l->str, l->str); string_append(&p, buf); } err = map(exports, "trust-info", 1, p, 0); @@ -1113,102 +1165,312 @@ struct http_response *trust_url(struct client_state *csp) } #endif /* def FEATURE_TRUST */ +/********************************************************************* + * + * Function : execute_single_pcrs_command + * + * Description : Apply single pcrs command to the subject. + * The subject itself is left untouched, memory for the result + * is malloc()ed and it is the caller's responsibility to free + * the result when it's no longer needed. + * + * Parameters : + * 1 : subject = the subject (== original) string + * 2 : pcrs_command = the pcrs command as string (s@foo@bar@) + * 3 : hits = int* for returning the number of modifications + * + * Returns : NULL in case of errors, otherwise the + * result of the pcrs command. + * + *********************************************************************/ +char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits) +{ + int error; + size_t size; + char *result = NULL; + pcrs_job *job; + + assert(subject); + assert(pcrs_command); + + *hits = 0; + size = strlen(subject); + + if (NULL == (job = pcrs_compile_command(pcrs_command, &error))) + { + log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.", + pcrs_command, error); + } + else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0) + { + log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits)); + *hits = 0; + freez(result); + } + + if (job) + { + job = pcrs_free_job(job); + } + + return result; + +} -#ifdef FEATURE_FAST_REDIRECTS /********************************************************************* * - * Function : redirect_url + * Function : rewrite_url * - * Description : Checks for redirection URLs and returns a HTTP redirect - * to the destination URL, if necessary + * Description : Rewrites a URL with a single pcrs command + * and returns the result if it differs from the + * original and isn't obviously invalid. * * Parameters : - * 1 : csp = Current client state (buffers, headers, etc...) + * 1 : old_url = URL to rewrite. + * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@) + * * - * Returns : NULL if URL was clean, HTTP redirect otherwise. + * Returns : NULL if the pcrs_command didn't change the url, or + * the result of the modification. * *********************************************************************/ -struct http_response *redirect_url(struct client_state *csp) +char *rewrite_url(char *old_url, const char *pcrs_command) { - char *p, *q; - struct http_response *rsp; - char *redirect_mode = NULL; - int x, y; + char *new_url = NULL; + int hits; - if ((csp->action->flags & ACTION_REDIRECT)) + assert(old_url); + assert(pcrs_command); + + new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits); + + if (hits == 0) { - q = csp->action->string[ACTION_STRING_REDIRECT]; + log_error(LOG_LEVEL_REDIRECTS, + "pcrs command \"%s\" didn't change \"%s\".", + pcrs_command, old_url, new_url); + freez(new_url); + } + else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8)) + { + log_error(LOG_LEVEL_ERROR, + "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), " + "but the result doesn't look like a valid URL and will be ignored.", + pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts"); + freez(new_url); } else { - redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS]; - if (0 == strcmpic(redirect_mode, "check-decoded-url")) - { - p = q = csp->http->path; - log_error(LOG_LEVEL_REDIRECTS, "Decoding path: %s if necessary.", p); - while (*p) - { - if (*p == '%') /* Escape sequence? */ - { - /* Yes, translate from hexadecimal to decimal */ - p++; - /* First byte */ - x=((int)*p++)-48; - if (x>9) x-=7; - x<<=4; - /* Second byte */ - y=((int)*p++)-48; - if (y>9)y-=7; - /* Merge */ - *q++=(char)(x|y); - } - else - { - /* No, forward character. */ - *q++=*p++; - } - } - *q='\0'; + log_error(LOG_LEVEL_REDIRECTS, + "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).", + pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts"); + } + + return new_url; + +} + + +#ifdef FEATURE_FAST_REDIRECTS +/********************************************************************* + * + * Function : get_last_url + * + * Description : Search for the last URL inside a string. + * If the string already is a URL, it will + * be the first URL found. + * + * Parameters : + * 1 : subject = the string to check + * 2 : redirect_mode = +fast-redirect{} mode + * + * Returns : NULL if no URL was found, or + * the last URL found. + * + *********************************************************************/ +char *get_last_url(char *subject, const char *redirect_mode) +{ + char *new_url = NULL; + char *tmp; + + assert(subject); + assert(redirect_mode); + + subject = strdup(subject); + if (subject == NULL) + { + log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects."); + return NULL; + } + + if (0 == strcmpic(redirect_mode, "check-decoded-url")) + { + log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject); + new_url = url_decode(subject); + if (new_url != NULL) + { + freez(subject); + subject = new_url; } - p = q = csp->http->path; - log_error(LOG_LEVEL_REDIRECTS, "Checking path for redirects: %s", p); + else + { + log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject); + } + } + + log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject); + + /* + * Find the last URL encoded in the request + */ + tmp = subject; + while ((tmp = strstr(tmp, "http://")) != NULL) + { + new_url = tmp++; + } + tmp = (new_url != NULL) ? new_url : subject; + while ((tmp = strstr(tmp, "https://")) != NULL) + { + new_url = tmp++; + } + if ((new_url != NULL) + && ( (new_url != subject) + || (0 == strncmpic(subject, "http://", 7)) + || (0 == strncmpic(subject, "https://", 8)) + )) + { /* - * find the last URL encoded in the request + * Return new URL if we found a redirect + * or if the subject already was a URL. + * + * The second case makes sure that we can + * chain get_last_url after another redirection check + * (like rewrite_url) without losing earlier redirects. */ - while ((p = strstr(p, "http://")) != NULL) - { - q = p++; - } + new_url = strdup(new_url); + freez(subject); + return new_url; } + + freez(subject); + return NULL; + +} +#endif /* def FEATURE_FAST_REDIRECTS */ + + +/********************************************************************* + * + * Function : redirect_url + * + * Description : Checks if Privoxy should answer the request with + * a HTTP redirect and generates the redirect if + * necessary. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * + * Returns : NULL if the request can pass, HTTP redirect otherwise. + * + *********************************************************************/ +struct http_response *redirect_url(struct client_state *csp) +{ + struct http_response *rsp; +#ifdef FEATURE_FAST_REDIRECTS /* - * if there was any, generate and return a HTTP redirect + * XXX: Do we still need FEATURE_FAST_REDIRECTS + * as compile-time option? The user can easily disable + * it in his action file. */ - if (q != csp->http->path) + char * redirect_mode; +#endif /* def FEATURE_FAST_REDIRECTS */ + char *old_url = NULL; + char *new_url = NULL; + char *redirection_string; + + if ((csp->action->flags & ACTION_REDIRECT)) { - log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q); + redirection_string = csp->action->string[ACTION_STRING_REDIRECT]; + + /* + * If the redirection string begins with 's', + * assume it's a pcrs command, otherwise treat it as + * properly formatted URL and use it for the redirection + * directly. + * + * According to RFC 2616 section 14.30 the URL + * has to be absolute and if the user tries: + * +redirect{shit/this/will/be/parsed/as/pcrs_command.html} + * she would get undefined results anyway. + * + */ - if (NULL == (rsp = alloc_http_response())) + if (*redirection_string == 's') { - return cgi_error_memory(); + old_url = csp->http->url; + new_url = rewrite_url(old_url, redirection_string); } - - if ( enlist_unique_header(rsp->headers, "Location", q) - || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) ) + else { - free_http_response(rsp); - return cgi_error_memory(); + log_error(LOG_LEVEL_REDIRECTS, + "No pcrs command recognized, assuming that \"%s\" is already properly formatted.", + redirection_string); + new_url = strdup(redirection_string); } + } - return finish_http_response(rsp); +#ifdef FEATURE_FAST_REDIRECTS + if ((csp->action->flags & ACTION_FAST_REDIRECTS)) + { + redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS]; + + /* + * If it exists, use the previously rewritten URL as input + * otherwise just use the old path. + */ + old_url = (new_url != NULL) ? new_url : strdup(csp->http->path); + new_url = get_last_url(old_url, redirect_mode); + freez(old_url); } - else +#endif /* def FEATURE_FAST_REDIRECTS */ + + /* Did any redirect action trigger? */ + if (new_url) { - return NULL; + if (0 == strcmpic(new_url, csp->http->url)) + { + log_error(LOG_LEVEL_ERROR, + "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.", + csp->http->url, new_url); + freez(new_url); + } + else + { + log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url); + + if (NULL == (rsp = alloc_http_response())) + { + freez(new_url); + return cgi_error_memory(); + } + + if ( enlist_unique_header(rsp->headers, "Location", new_url) + || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) ) + { + freez(new_url); + free_http_response(rsp); + return cgi_error_memory(); + } + freez(new_url); + return finish_http_response(rsp); + } } + /* Only reached if no redirect is required */ + return NULL; + } -#endif /* def FEATURE_FAST_REDIRECTS */ #ifdef FEATURE_IMAGE_BLOCKING @@ -1229,7 +1491,7 @@ struct http_response *redirect_url(struct client_state *csp) * otherwise * *********************************************************************/ -int is_imageurl(struct client_state *csp) +int is_imageurl(const struct client_state *csp) { #ifdef FEATURE_IMAGE_DETECT_MSIE char *tmp; @@ -1276,7 +1538,7 @@ int is_imageurl(struct client_state *csp) * Returns : 0 => trusted, 1 => untrusted * *********************************************************************/ -int is_untrusted_url(struct client_state *csp) +int is_untrusted_url(const struct client_state *csp) { struct file_list *fl; struct block_spec *b; @@ -1328,7 +1590,7 @@ int is_untrusted_url(struct client_state *csp) { /* if the URL's referrer is from a trusted referrer, then * add the target spec to the trustfile as an unblocked - * domain and return NULL (which means it's OK). + * domain and return 0 (which means it's OK). */ FILE *fp; @@ -1359,9 +1621,19 @@ int is_untrusted_url(struct client_state *csp) string_join(&new_entry, path); } + /* + * Give a reason for generating this entry. + */ + string_append(&new_entry, " # Trusted referrer was: "); + string_append(&new_entry, referer); + if (new_entry != NULL) { - fprintf(fp, "%s\n", new_entry); + if (-1 == fprintf(fp, "%s\n", new_entry)) + { + log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E", + new_entry, csp->config->trustfile); + } free(new_entry); } else @@ -1372,6 +1644,11 @@ int is_untrusted_url(struct client_state *csp) fclose(fp); } + else + { + log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E", + csp->http->hostport, csp->config->trustfile); + } return 0; } } @@ -1390,6 +1667,11 @@ int is_untrusted_url(struct client_state *csp) * csp->content_length to the modified size and raise the * CSP_FLAG_MODIFIED flag. * + * XXX: Currently pcrs_filter_response is also responsible + * for dechunking and decompressing. Both should be + * done in separate functions so other content modifiers + * profit as well, even if pcrs filtering is disabled. + * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * @@ -1418,7 +1700,7 @@ char *pcrs_filter_response(struct client_state *csp) { return(NULL); } - size = csp->iob->eod - csp->iob->cur; + size = (size_t)(csp->iob->eod - csp->iob->cur); /* * Need to check the set of re_filterfiles... @@ -1439,7 +1721,7 @@ char *pcrs_filter_response(struct client_state *csp) if (0 == found_filters) { log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering."); - return(NULL); + return(NULL); } /* @@ -1457,11 +1739,57 @@ char *pcrs_filter_response(struct client_state *csp) csp->flags |= CSP_FLAG_MODIFIED; } +#ifdef FEATURE_ZLIB + /* + * If the body has a compressed transfer-encoding, + * uncompress it first, adjusting size and iob->eod. + * Note that decompression occurs after de-chunking. + */ + if (csp->content_type & CT_GZIP || csp->content_type & CT_DEFLATE) + { + /* Notice that we at least tried to decompress. */ + if (JB_ERR_OK != decompress_iob(csp)) + { + /* + * We failed to decompress the data; there's no point + * in continuing since we can't filter. This is + * slightly tricky because we need to remember not to + * modify the Content-Encoding header later; using + * CT_TABOO flag is a kludge for this purpose. + */ + csp->content_type |= CT_TABOO; + return(NULL); + } + log_error(LOG_LEVEL_RE_FILTER, "Decompressing successful"); + + /* + * Decompression gives us a completely new iob, + * so we need to update. + */ + size = (size_t)(csp->iob->eod - csp->iob->cur); + old = csp->iob->cur; + + csp->flags |= CSP_FLAG_MODIFIED; + csp->content_type &= ~CT_TABOO; + } +#endif + for (i = 0; i < MAX_AF_FILES; i++) { fl = csp->rlist[i]; if ((NULL == fl) || (NULL == fl->f)) - break; + { + /* + * Either there are no filter files + * left, or this filter file just + * contains no valid filters. + * + * Continue to be sure we don't miss + * valid filter files that are chained + * after empty or invalid ones. + */ + continue; + } /* * For all applying +filter actions, look if a filter by that * name exists and if yes, execute it's pcrs_joblist on the @@ -1474,7 +1802,9 @@ char *pcrs_filter_response(struct client_state *csp) { if (strcmp(b->name, filtername->str) == 0) { - int current_hits = 0; + int current_hits = 0; /* Number of hits caused by this filter */ + int job_number = 0; /* Which job we're currently executing */ + int job_hits = 0; /* How many hits the current job caused */ if ( NULL == b->joblist ) { @@ -1486,13 +1816,50 @@ char *pcrs_filter_response(struct client_state *csp) /* Apply all jobs from the joblist */ for (job = b->joblist; NULL != job; job = job->next) { - current_hits += pcrs_execute(job, old, size, &new, &size); - if (old != csp->iob->cur) free(old); - old=new; + job_number++; + job_hits = pcrs_execute(job, old, size, &new, &size); + + if (job_hits >= 0) + { + /* + * That went well. Continue filtering + * and use the result of this job as + * input for the next one. + */ + current_hits += job_hits; + if (old != csp->iob->cur) + { + free(old); + } + old = new; + } + else + { + /* + * This job caused an unexpected error. Inform the user + * and skip the rest of the jobs in this filter. We could + * continue with the next job, but usually the jobs + * depend on each other or are similar enough to + * fail for the same reason. + * + * At the moment our pcrs expects the error codes of pcre 3.4, + * but newer pcre versions can return additional error codes. + * As a result pcrs_strerror()'s error message might be + * "Unknown error ...", therefore we print the numerical value + * as well. + * + * XXX: Is this important enough for LOG_LEVEL_ERROR or + * should we use LOG_LEVEL_RE_FILTER instead? + */ + log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)", + b->name, job_number, pcrs_strerror(job_hits), job_hits); + break; + } } - log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).", - csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size); + log_error(LOG_LEVEL_RE_FILTER, + "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).", + csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size); hits += current_hits; } @@ -1538,7 +1905,9 @@ char *gif_deanimate_response(struct client_state *csp) { struct binbuffer *in, *out; char *p; - size_t size = csp->iob->eod - csp->iob->cur; + size_t size; + + size = (size_t)(csp->iob->eod - csp->iob->cur); /* * If the body has a "chunked" transfer-encoding, @@ -1608,9 +1977,12 @@ char *gif_deanimate_response(struct client_state *csp) *********************************************************************/ char *jpeg_inspect_response(struct client_state *csp) { - struct binbuffer *in = NULL, *out = NULL; + struct binbuffer *in = NULL; + struct binbuffer *out = NULL; char *p = NULL; - size_t size = csp->iob->eod - csp->iob->cur; + size_t size; + + size = (size_t)(csp->iob->eod - csp->iob->cur); /* * If the body has a "chunked" transfer-encoding, @@ -1683,7 +2055,7 @@ char *jpeg_inspect_response(struct client_state *csp) * went wrong * *********************************************************************/ -int remove_chunked_transfer_coding(char *buffer, const size_t size) +size_t remove_chunked_transfer_coding(char *buffer, const size_t size) { size_t newsize = 0; unsigned int chunksize = 0;