1 /* Licensed to the Apache Software Foundation (ASF) under one or more
2 * contributor license agreements. See the NOTICE file distributed with
3 * this work for additional information regarding copyright ownership.
4 * The ASF licenses this file to You under the Apache License, Version 2.0
5 * (the "License"); you may not use this file except in compliance with
6 * the License. You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
17 /* HTTP routines for Apache proxy */
19 #include "mod_proxy.h"
22 module AP_MODULE_DECLARE_DATA proxy_http_module;
24 static apr_status_t ap_proxy_http_cleanup(const char *scheme,
26 proxy_conn_rec *backend);
29 * Canonicalise http-like URLs.
30 * scheme is the scheme for the URL
31 * url is the URL starting with the first '/'
32 * def_port is the default port for this scheme.
34 static int proxy_http_canon(request_rec *r, char *url)
36 char *host, *path, sport[7];
40 apr_port_t port, def_port;
42 /* ap_port_of_scheme() */
43 if (strncasecmp(url, "http:", 5) == 0) {
47 else if (strncasecmp(url, "https:", 6) == 0) {
54 def_port = apr_uri_port_of_scheme(scheme);
56 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
57 "proxy: HTTP: canonicalising URL %s", url);
60 * We break the URL into host, port, path, search
63 err = ap_proxy_canon_netloc(r->pool, &url, NULL, NULL, &host, &port);
65 ap_log_rerror(APLOG_MARK, APLOG_ERR, 0, r,
66 "error parsing URL %s: %s",
68 return HTTP_BAD_REQUEST;
72 * now parse path/search args, according to rfc1738:
75 * In a reverse proxy, our URL has been processed, so canonicalise
76 * unless proxy-nocanon is set to say it's raw
77 * In a forward proxy, we have and MUST NOT MANGLE the original.
79 switch (r->proxyreq) {
80 default: /* wtf are we doing here? */
81 case PROXYREQ_REVERSE:
82 if (apr_table_get(r->notes, "proxy-nocanon")) {
83 path = url; /* this is the raw path */
86 path = ap_proxy_canonenc(r->pool, url, strlen(url),
87 enc_path, 0, r->proxyreq);
97 return HTTP_BAD_REQUEST;
100 apr_snprintf(sport, sizeof(sport), ":%d", port);
104 if (ap_strchr_c(host, ':')) { /* if literal IPv6 address */
105 host = apr_pstrcat(r->pool, "[", host, "]", NULL);
107 r->filename = apr_pstrcat(r->pool, "proxy:", scheme, "://", host, sport,
108 "/", path, (search) ? "?" : "", (search) ? search : "", NULL);
112 /* Clear all connection-based headers from the incoming headers table */
113 typedef struct header_dptr {
118 static ap_regex_t *warn_rx;
119 static int clean_warning_headers(void *data, const char *key, const char *val)
121 apr_table_t *headers = ((header_dptr*)data)->table;
122 apr_pool_t *pool = ((header_dptr*)data)->pool;
125 apr_time_t warn_time;
126 const int nmatch = 3;
127 ap_regmatch_t pmatch[3];
129 if (headers == NULL) {
130 ((header_dptr*)data)->table = headers = apr_table_make(pool, 2);
133 * Parse this, suckers!
135 * Warning = "Warning" ":" 1#warning-value
137 * warning-value = warn-code SP warn-agent SP warn-text
141 * warn-agent = ( host [ ":" port ] ) | pseudonym
142 * ; the name or pseudonym of the server adding
143 * ; the Warning header, for use in debugging
144 * warn-text = quoted-string
145 * warn-date = <"> HTTP-date <">
147 * Buggrit, use a bloomin' regexp!
148 * (\d{3}\s+\S+\s+\".*?\"(\s+\"(.*?)\")?) --> whole in $1, date in $3
150 while (!ap_regexec(warn_rx, val, nmatch, pmatch, 0)) {
151 warning = apr_pstrndup(pool, val+pmatch[0].rm_so,
152 pmatch[0].rm_eo - pmatch[0].rm_so);
154 if (pmatch[2].rm_eo > pmatch[2].rm_so) {
155 /* OK, we have a date here */
156 date = apr_pstrndup(pool, val+pmatch[2].rm_so,
157 pmatch[2].rm_eo - pmatch[2].rm_so);
158 warn_time = apr_date_parse_http(date);
160 if (!warn_time || (warn_time == ((header_dptr*)data)->time)) {
161 apr_table_addn(headers, key, warning);
163 val += pmatch[0].rm_eo;
167 static apr_table_t *ap_proxy_clean_warnings(apr_pool_t *p, apr_table_t *headers)
172 x.time = apr_date_parse_http(apr_table_get(headers, "Date"));
173 apr_table_do(clean_warning_headers, &x, headers, "Warning", NULL);
174 if (x.table != NULL) {
175 apr_table_unset(headers, "Warning");
176 return apr_table_overlay(p, headers, x.table);
182 static int clear_conn_headers(void *data, const char *key, const char *val)
184 apr_table_t *headers = ((header_dptr*)data)->table;
185 apr_pool_t *pool = ((header_dptr*)data)->pool;
187 char *next = apr_pstrdup(pool, val);
190 while (*next && !apr_isspace(*next) && (*next != ',')) {
193 while (*next && (apr_isspace(*next) || (*next == ','))) {
196 apr_table_unset(headers, name);
200 static void ap_proxy_clear_connection(apr_pool_t *p, apr_table_t *headers)
205 apr_table_unset(headers, "Proxy-Connection");
206 apr_table_do(clear_conn_headers, &x, headers, "Connection", NULL);
207 apr_table_unset(headers, "Connection");
209 static void add_te_chunked(apr_pool_t *p,
210 apr_bucket_alloc_t *bucket_alloc,
211 apr_bucket_brigade *header_brigade)
215 const char te_hdr[] = "Transfer-Encoding: chunked" CRLF;
217 buf = apr_pmemdup(p, te_hdr, sizeof(te_hdr)-1);
218 ap_xlate_proto_to_ascii(buf, sizeof(te_hdr)-1);
220 e = apr_bucket_pool_create(buf, sizeof(te_hdr)-1, p, bucket_alloc);
221 APR_BRIGADE_INSERT_TAIL(header_brigade, e);
224 static void add_cl(apr_pool_t *p,
225 apr_bucket_alloc_t *bucket_alloc,
226 apr_bucket_brigade *header_brigade,
232 buf = apr_pstrcat(p, "Content-Length: ",
236 ap_xlate_proto_to_ascii(buf, strlen(buf));
237 e = apr_bucket_pool_create(buf, strlen(buf), p, bucket_alloc);
238 APR_BRIGADE_INSERT_TAIL(header_brigade, e);
241 #define ASCII_CRLF "\015\012"
242 #define ASCII_ZERO "\060"
244 static void terminate_headers(apr_bucket_alloc_t *bucket_alloc,
245 apr_bucket_brigade *header_brigade)
249 /* add empty line at the end of the headers */
250 e = apr_bucket_immortal_create(ASCII_CRLF, 2, bucket_alloc);
251 APR_BRIGADE_INSERT_TAIL(header_brigade, e);
254 static int pass_brigade(apr_bucket_alloc_t *bucket_alloc,
255 request_rec *r, proxy_conn_rec *conn,
256 conn_rec *origin, apr_bucket_brigade *bb,
260 apr_off_t transferred;
263 apr_bucket *e = apr_bucket_flush_create(bucket_alloc);
264 APR_BRIGADE_INSERT_TAIL(bb, e);
266 apr_brigade_length(bb, 0, &transferred);
267 if (transferred != -1)
268 conn->worker->s->transferred += transferred;
269 status = ap_pass_brigade(origin->output_filters, bb);
270 if (status != APR_SUCCESS) {
271 ap_log_error(APLOG_MARK, APLOG_ERR, status, r->server,
272 "proxy: pass request body failed to %pI (%s)",
273 conn->addr, conn->hostname);
274 if (origin->aborted) {
275 return APR_STATUS_IS_TIMEUP(status) ? HTTP_GATEWAY_TIME_OUT : HTTP_BAD_GATEWAY;
278 return HTTP_BAD_REQUEST;
281 apr_brigade_cleanup(bb);
285 #define MAX_MEM_SPOOL 16384
287 static int stream_reqbody_chunked(apr_pool_t *p,
289 proxy_conn_rec *p_conn,
291 apr_bucket_brigade *header_brigade,
292 apr_bucket_brigade *input_brigade)
294 int seen_eos = 0, rv = OK;
298 apr_bucket_alloc_t *bucket_alloc = r->connection->bucket_alloc;
299 apr_bucket_brigade *bb;
302 add_te_chunked(p, bucket_alloc, header_brigade);
303 terminate_headers(bucket_alloc, header_brigade);
305 while (!APR_BUCKET_IS_EOS(APR_BRIGADE_FIRST(input_brigade)))
307 char chunk_hdr[20]; /* must be here due to transient bucket. */
309 /* If this brigade contains EOS, either stop or remove it. */
310 if (APR_BUCKET_IS_EOS(APR_BRIGADE_LAST(input_brigade))) {
313 /* We can't pass this EOS to the output_filters. */
314 e = APR_BRIGADE_LAST(input_brigade);
315 apr_bucket_delete(e);
318 apr_brigade_length(input_brigade, 1, &bytes);
320 hdr_len = apr_snprintf(chunk_hdr, sizeof(chunk_hdr),
321 "%" APR_UINT64_T_HEX_FMT CRLF,
322 (apr_uint64_t)bytes);
324 ap_xlate_proto_to_ascii(chunk_hdr, hdr_len);
325 e = apr_bucket_transient_create(chunk_hdr, hdr_len,
327 APR_BRIGADE_INSERT_HEAD(input_brigade, e);
330 * Append the end-of-chunk CRLF
332 e = apr_bucket_immortal_create(ASCII_CRLF, 2, bucket_alloc);
333 APR_BRIGADE_INSERT_TAIL(input_brigade, e);
335 if (header_brigade) {
336 /* we never sent the header brigade, so go ahead and
337 * take care of that now
342 * Save input_brigade in bb brigade. (At least) in the SSL case
343 * input_brigade contains transient buckets whose data would get
344 * overwritten during the next call of ap_get_brigade in the loop.
345 * ap_save_brigade ensures these buckets to be set aside.
346 * Calling ap_save_brigade with NULL as filter is OK, because
347 * bb brigade already has been created and does not need to get
348 * created by ap_save_brigade.
350 status = ap_save_brigade(NULL, &bb, &input_brigade, p);
351 if (status != APR_SUCCESS) {
352 return HTTP_INTERNAL_SERVER_ERROR;
355 header_brigade = NULL;
361 /* The request is flushed below this loop with chunk EOS header */
362 rv = pass_brigade(bucket_alloc, r, p_conn, origin, bb, 0);
371 status = ap_get_brigade(r->input_filters, input_brigade,
372 AP_MODE_READBYTES, APR_BLOCK_READ,
375 if (status != APR_SUCCESS) {
376 return HTTP_BAD_REQUEST;
380 if (header_brigade) {
381 /* we never sent the header brigade because there was no request body;
387 if (!APR_BRIGADE_EMPTY(input_brigade)) {
388 /* input brigade still has an EOS which we can't pass to the output_filters. */
389 e = APR_BRIGADE_LAST(input_brigade);
390 AP_DEBUG_ASSERT(APR_BUCKET_IS_EOS(e));
391 apr_bucket_delete(e);
396 e = apr_bucket_immortal_create(ASCII_ZERO ASCII_CRLF
400 APR_BRIGADE_INSERT_TAIL(bb, e);
402 if (apr_table_get(r->subprocess_env, "proxy-sendextracrlf")) {
403 e = apr_bucket_immortal_create(ASCII_CRLF, 2, bucket_alloc);
404 APR_BRIGADE_INSERT_TAIL(bb, e);
407 /* Now we have headers-only, or the chunk EOS mark; flush it */
408 rv = pass_brigade(bucket_alloc, r, p_conn, origin, bb, 1);
412 static int stream_reqbody_cl(apr_pool_t *p,
414 proxy_conn_rec *p_conn,
416 apr_bucket_brigade *header_brigade,
417 apr_bucket_brigade *input_brigade,
418 const char *old_cl_val)
420 int seen_eos = 0, rv = 0;
421 apr_status_t status = APR_SUCCESS;
422 apr_bucket_alloc_t *bucket_alloc = r->connection->bucket_alloc;
423 apr_bucket_brigade *bb;
425 apr_off_t cl_val = 0;
427 apr_off_t bytes_streamed = 0;
430 add_cl(p, bucket_alloc, header_brigade, old_cl_val);
431 if (APR_SUCCESS != (status = apr_strtoff(&cl_val, old_cl_val, NULL,
433 return HTTP_INTERNAL_SERVER_ERROR;
436 terminate_headers(bucket_alloc, header_brigade);
438 while (!APR_BUCKET_IS_EOS(APR_BRIGADE_FIRST(input_brigade)))
440 apr_brigade_length(input_brigade, 1, &bytes);
441 bytes_streamed += bytes;
443 /* If this brigade contains EOS, either stop or remove it. */
444 if (APR_BUCKET_IS_EOS(APR_BRIGADE_LAST(input_brigade))) {
447 /* We can't pass this EOS to the output_filters. */
448 e = APR_BRIGADE_LAST(input_brigade);
449 apr_bucket_delete(e);
451 if (apr_table_get(r->subprocess_env, "proxy-sendextracrlf")) {
452 e = apr_bucket_immortal_create(ASCII_CRLF, 2, bucket_alloc);
453 APR_BRIGADE_INSERT_TAIL(input_brigade, e);
457 /* C-L < bytes streamed?!?
458 * We will error out after the body is completely
459 * consumed, but we can't stream more bytes at the
460 * back end since they would in part be interpreted
461 * as another request! If nothing is sent, then
464 * Prevents HTTP Response Splitting.
466 if (bytes_streamed > cl_val)
469 if (header_brigade) {
470 /* we never sent the header brigade, so go ahead and
471 * take care of that now
476 * Save input_brigade in bb brigade. (At least) in the SSL case
477 * input_brigade contains transient buckets whose data would get
478 * overwritten during the next call of ap_get_brigade in the loop.
479 * ap_save_brigade ensures these buckets to be set aside.
480 * Calling ap_save_brigade with NULL as filter is OK, because
481 * bb brigade already has been created and does not need to get
482 * created by ap_save_brigade.
484 status = ap_save_brigade(NULL, &bb, &input_brigade, p);
485 if (status != APR_SUCCESS) {
486 return HTTP_INTERNAL_SERVER_ERROR;
489 header_brigade = NULL;
495 /* Once we hit EOS, we are ready to flush. */
496 rv = pass_brigade(bucket_alloc, r, p_conn, origin, bb, seen_eos);
505 status = ap_get_brigade(r->input_filters, input_brigade,
506 AP_MODE_READBYTES, APR_BLOCK_READ,
509 if (status != APR_SUCCESS) {
510 return HTTP_BAD_REQUEST;
514 if (bytes_streamed != cl_val) {
515 ap_log_error(APLOG_MARK, APLOG_ERR, 0, r->server,
516 "proxy: client %s given Content-Length did not match"
517 " number of body bytes read", r->connection->remote_ip);
518 return HTTP_BAD_REQUEST;
521 if (header_brigade) {
522 /* we never sent the header brigade since there was no request
523 * body; send it now with the flush flag
526 return(pass_brigade(bucket_alloc, r, p_conn, origin, bb, 1));
532 static int spool_reqbody_cl(apr_pool_t *p,
534 proxy_conn_rec *p_conn,
536 apr_bucket_brigade *header_brigade,
537 apr_bucket_brigade *input_brigade,
542 apr_bucket_alloc_t *bucket_alloc = r->connection->bucket_alloc;
543 apr_bucket_brigade *body_brigade;
545 apr_off_t bytes, bytes_spooled = 0, fsize = 0;
546 apr_file_t *tmpfile = NULL;
548 body_brigade = apr_brigade_create(p, bucket_alloc);
550 while (!APR_BUCKET_IS_EOS(APR_BRIGADE_FIRST(input_brigade)))
552 /* If this brigade contains EOS, either stop or remove it. */
553 if (APR_BUCKET_IS_EOS(APR_BRIGADE_LAST(input_brigade))) {
556 /* We can't pass this EOS to the output_filters. */
557 e = APR_BRIGADE_LAST(input_brigade);
558 apr_bucket_delete(e);
561 apr_brigade_length(input_brigade, 1, &bytes);
563 if (bytes_spooled + bytes > MAX_MEM_SPOOL) {
564 /* can't spool any more in memory; write latest brigade to disk */
565 if (tmpfile == NULL) {
566 const char *temp_dir;
569 status = apr_temp_dir_get(&temp_dir, p);
570 if (status != APR_SUCCESS) {
571 ap_log_error(APLOG_MARK, APLOG_ERR, status, r->server,
572 "proxy: search for temporary directory failed");
573 return HTTP_INTERNAL_SERVER_ERROR;
575 apr_filepath_merge(&template, temp_dir,
576 "modproxy.tmp.XXXXXX",
577 APR_FILEPATH_NATIVE, p);
578 status = apr_file_mktemp(&tmpfile, template, 0, p);
579 if (status != APR_SUCCESS) {
580 ap_log_error(APLOG_MARK, APLOG_ERR, status, r->server,
581 "proxy: creation of temporary file in directory %s failed",
583 return HTTP_INTERNAL_SERVER_ERROR;
586 for (e = APR_BRIGADE_FIRST(input_brigade);
587 e != APR_BRIGADE_SENTINEL(input_brigade);
588 e = APR_BUCKET_NEXT(e)) {
590 apr_size_t bytes_read, bytes_written;
592 apr_bucket_read(e, &data, &bytes_read, APR_BLOCK_READ);
593 status = apr_file_write_full(tmpfile, data, bytes_read, &bytes_written);
594 if (status != APR_SUCCESS) {
595 const char *tmpfile_name;
597 if (apr_file_name_get(&tmpfile_name, tmpfile) != APR_SUCCESS) {
598 tmpfile_name = "(unknown)";
600 ap_log_error(APLOG_MARK, APLOG_ERR, status, r->server,
601 "proxy: write to temporary file %s failed",
603 return HTTP_INTERNAL_SERVER_ERROR;
605 AP_DEBUG_ASSERT(bytes_read == bytes_written);
606 fsize += bytes_written;
608 apr_brigade_cleanup(input_brigade);
613 * Save input_brigade in body_brigade. (At least) in the SSL case
614 * input_brigade contains transient buckets whose data would get
615 * overwritten during the next call of ap_get_brigade in the loop.
616 * ap_save_brigade ensures these buckets to be set aside.
617 * Calling ap_save_brigade with NULL as filter is OK, because
618 * body_brigade already has been created and does not need to get
619 * created by ap_save_brigade.
621 status = ap_save_brigade(NULL, &body_brigade, &input_brigade, p);
622 if (status != APR_SUCCESS) {
623 return HTTP_INTERNAL_SERVER_ERROR;
628 bytes_spooled += bytes;
634 status = ap_get_brigade(r->input_filters, input_brigade,
635 AP_MODE_READBYTES, APR_BLOCK_READ,
638 if (status != APR_SUCCESS) {
639 return HTTP_BAD_REQUEST;
643 if (bytes_spooled || force_cl) {
644 add_cl(p, bucket_alloc, header_brigade, apr_off_t_toa(p, bytes_spooled));
646 terminate_headers(bucket_alloc, header_brigade);
647 APR_BRIGADE_CONCAT(header_brigade, body_brigade);
649 apr_brigade_insert_file(header_brigade, tmpfile, 0, fsize, p);
651 if (apr_table_get(r->subprocess_env, "proxy-sendextracrlf")) {
652 e = apr_bucket_immortal_create(ASCII_CRLF, 2, bucket_alloc);
653 APR_BRIGADE_INSERT_TAIL(header_brigade, e);
655 /* This is all a single brigade, pass with flush flagged */
656 return(pass_brigade(bucket_alloc, r, p_conn, origin, header_brigade, 1));
660 int ap_proxy_http_request(apr_pool_t *p, request_rec *r,
661 proxy_conn_rec *p_conn, conn_rec *origin,
662 proxy_server_conf *conf,
664 char *url, char *server_portstr)
666 conn_rec *c = r->connection;
667 apr_bucket_alloc_t *bucket_alloc = c->bucket_alloc;
668 apr_bucket_brigade *header_brigade;
669 apr_bucket_brigade *input_brigade;
670 apr_bucket_brigade *temp_brigade;
673 const apr_array_header_t *headers_in_array;
674 const apr_table_entry_t *headers_in;
677 enum rb_methods {RB_INIT, RB_STREAM_CL, RB_STREAM_CHUNKED, RB_SPOOL_CL};
678 enum rb_methods rb_method = RB_INIT;
679 const char *old_cl_val = NULL;
680 const char *old_te_val = NULL;
681 apr_off_t bytes_read = 0;
684 apr_table_t *headers_in_copy;
686 header_brigade = apr_brigade_create(p, origin->bucket_alloc);
689 * Send the HTTP/1.1 request to the remote server
692 if (apr_table_get(r->subprocess_env, "force-proxy-request-1.0")) {
694 * According to RFC 2616 8.2.3 we are not allowed to forward an
695 * Expect: 100-continue to an HTTP/1.0 server. Instead we MUST return
696 * a HTTP_EXPECTATION_FAILED
698 if (r->expecting_100) {
699 return HTTP_EXPECTATION_FAILED;
701 buf = apr_pstrcat(p, r->method, " ", url, " HTTP/1.0" CRLF, NULL);
705 buf = apr_pstrcat(p, r->method, " ", url, " HTTP/1.1" CRLF, NULL);
708 if (apr_table_get(r->subprocess_env, "proxy-nokeepalive")) {
709 origin->keepalive = AP_CONN_CLOSE;
712 ap_xlate_proto_to_ascii(buf, strlen(buf));
713 e = apr_bucket_pool_create(buf, strlen(buf), p, c->bucket_alloc);
714 APR_BRIGADE_INSERT_TAIL(header_brigade, e);
715 if (conf->preserve_host == 0) {
716 if (uri->port_str && uri->port != DEFAULT_HTTP_PORT) {
717 buf = apr_pstrcat(p, "Host: ", uri->hostname, ":", uri->port_str,
720 buf = apr_pstrcat(p, "Host: ", uri->hostname, CRLF, NULL);
724 /* don't want to use r->hostname, as the incoming header might have a
727 const char* hostname = apr_table_get(r->headers_in,"Host");
729 hostname = r->server->server_hostname;
730 ap_log_rerror(APLOG_MARK, APLOG_WARNING, 0, r,
731 "proxy: no HTTP 0.9 request (with no host line) "
732 "on incoming request and preserve host set "
733 "forcing hostname to be %s for uri %s",
737 buf = apr_pstrcat(p, "Host: ", hostname, CRLF, NULL);
739 ap_xlate_proto_to_ascii(buf, strlen(buf));
740 e = apr_bucket_pool_create(buf, strlen(buf), p, c->bucket_alloc);
741 APR_BRIGADE_INSERT_TAIL(header_brigade, e);
744 if (conf->viaopt == via_block) {
745 /* Block all outgoing Via: headers */
746 apr_table_unset(r->headers_in, "Via");
747 } else if (conf->viaopt != via_off) {
748 const char *server_name = ap_get_server_name(r);
749 /* If USE_CANONICAL_NAME_OFF was configured for the proxy virtual host,
750 * then the server name returned by ap_get_server_name() is the
751 * origin server name (which does make too much sense with Via: headers)
752 * so we use the proxy vhost's name instead.
754 if (server_name == r->hostname)
755 server_name = r->server->server_hostname;
756 /* Create a "Via:" request header entry and merge it */
757 /* Generate outgoing Via: header with/without server comment: */
758 apr_table_mergen(r->headers_in, "Via",
759 (conf->viaopt == via_full)
760 ? apr_psprintf(p, "%d.%d %s%s (%s)",
761 HTTP_VERSION_MAJOR(r->proto_num),
762 HTTP_VERSION_MINOR(r->proto_num),
763 server_name, server_portstr,
764 AP_SERVER_BASEVERSION)
765 : apr_psprintf(p, "%d.%d %s%s",
766 HTTP_VERSION_MAJOR(r->proto_num),
767 HTTP_VERSION_MINOR(r->proto_num),
768 server_name, server_portstr)
772 /* X-Forwarded-*: handling
777 * These request headers are only really useful when the mod_proxy
778 * is used in a reverse proxy configuration, so that useful info
779 * about the client can be passed through the reverse proxy and on
780 * to the backend server, which may require the information to
783 * In a forward proxy situation, these options are a potential
784 * privacy violation, as information about clients behind the proxy
785 * are revealed to arbitrary servers out there on the internet.
787 * The HTTP/1.1 Via: header is designed for passing client
788 * information through proxies to a server, and should be used in
789 * a forward proxy configuation instead of X-Forwarded-*. See the
790 * ProxyVia option for details.
793 if (PROXYREQ_REVERSE == r->proxyreq) {
796 /* Add X-Forwarded-For: so that the upstream has a chance to
797 * determine, where the original request came from.
799 apr_table_mergen(r->headers_in, "X-Forwarded-For",
802 /* Add X-Forwarded-Host: so that upstream knows what the
803 * original request hostname was.
805 if ((buf = apr_table_get(r->headers_in, "Host"))) {
806 apr_table_mergen(r->headers_in, "X-Forwarded-Host", buf);
809 /* Add X-Forwarded-Server: so that upstream knows what the
810 * name of this proxy server is (if there are more than one)
811 * XXX: This duplicates Via: - do we strictly need it?
813 apr_table_mergen(r->headers_in, "X-Forwarded-Server",
814 r->server->server_hostname);
819 * Make a copy of the headers_in table before clearing the connection
820 * headers as we need the connection headers later in the http output
821 * filter to prepare the correct response headers.
823 * Note: We need to take r->pool for apr_table_copy as the key / value
824 * pairs in r->headers_in have been created out of r->pool and
825 * p might be (and actually is) a longer living pool.
826 * This would trigger the bad pool ancestry abort in apr_table_copy if
827 * apr is compiled with APR_POOL_DEBUG.
829 headers_in_copy = apr_table_copy(r->pool, r->headers_in);
830 ap_proxy_clear_connection(p, headers_in_copy);
831 /* send request headers */
832 headers_in_array = apr_table_elts(headers_in_copy);
833 headers_in = (const apr_table_entry_t *) headers_in_array->elts;
834 for (counter = 0; counter < headers_in_array->nelts; counter++) {
835 if (headers_in[counter].key == NULL
836 || headers_in[counter].val == NULL
839 || !strcasecmp(headers_in[counter].key, "Host")
841 /* Clear out hop-by-hop request headers not to send
842 * RFC2616 13.5.1 says we should strip these headers
844 || !strcasecmp(headers_in[counter].key, "Keep-Alive")
845 || !strcasecmp(headers_in[counter].key, "TE")
846 || !strcasecmp(headers_in[counter].key, "Trailer")
847 || !strcasecmp(headers_in[counter].key, "Upgrade")
852 /* Do we want to strip Proxy-Authorization ?
853 * If we haven't used it, then NO
854 * If we have used it then MAYBE: RFC2616 says we MAY propagate it.
855 * So let's make it configurable by env.
857 if (!strcasecmp(headers_in[counter].key,"Proxy-Authorization")) {
858 if (r->user != NULL) { /* we've authenticated */
859 if (!apr_table_get(r->subprocess_env, "Proxy-Chain-Auth")) {
866 /* Skip Transfer-Encoding and Content-Length for now.
868 if (!strcasecmp(headers_in[counter].key, "Transfer-Encoding")) {
869 old_te_val = headers_in[counter].val;
872 if (!strcasecmp(headers_in[counter].key, "Content-Length")) {
873 old_cl_val = headers_in[counter].val;
877 /* for sub-requests, ignore freshness/expiry headers */
879 if ( !strcasecmp(headers_in[counter].key, "If-Match")
880 || !strcasecmp(headers_in[counter].key, "If-Modified-Since")
881 || !strcasecmp(headers_in[counter].key, "If-Range")
882 || !strcasecmp(headers_in[counter].key, "If-Unmodified-Since")
883 || !strcasecmp(headers_in[counter].key, "If-None-Match")) {
888 buf = apr_pstrcat(p, headers_in[counter].key, ": ",
889 headers_in[counter].val, CRLF,
891 ap_xlate_proto_to_ascii(buf, strlen(buf));
892 e = apr_bucket_pool_create(buf, strlen(buf), p, c->bucket_alloc);
893 APR_BRIGADE_INSERT_TAIL(header_brigade, e);
896 /* We have headers, let's figure out our request body... */
897 input_brigade = apr_brigade_create(p, bucket_alloc);
899 /* sub-requests never use keepalives, and mustn't pass request bodies.
900 * Because the new logic looks at input_brigade, we will self-terminate
901 * input_brigade and jump past all of the request body logic...
902 * Reading anything with ap_get_brigade is likely to consume the
903 * main request's body or read beyond EOS - which would be unplesant.
905 * An exception: when a kept_body is present, then subrequest CAN use
906 * pass request bodies, and we DONT skip the body.
908 if (!r->kept_body && r->main) {
909 /* XXX: Why DON'T sub-requests use keepalives? */
913 apr_table_unset(r->headers_in, "Content-Length");
917 apr_table_unset(r->headers_in, "Transfer-Encoding");
919 rb_method = RB_STREAM_CL;
920 e = apr_bucket_eos_create(input_brigade->bucket_alloc);
921 APR_BRIGADE_INSERT_TAIL(input_brigade, e);
925 /* WE only understand chunked. Other modules might inject
926 * (and therefore, decode) other flavors but we don't know
927 * that the can and have done so unless they they remove
928 * their decoding from the headers_in T-E list.
929 * XXX: Make this extensible, but in doing so, presume the
930 * encoding has been done by the extensions' handler, and
931 * do not modify add_te_chunked's logic
933 if (old_te_val && strcmp(old_te_val, "chunked") != 0) {
934 ap_log_error(APLOG_MARK, APLOG_ERR, 0, r->server,
935 "proxy: %s Transfer-Encoding is not supported",
937 return HTTP_INTERNAL_SERVER_ERROR;
940 if (old_cl_val && old_te_val) {
941 ap_log_error(APLOG_MARK, APLOG_DEBUG, APR_ENOTIMPL, r->server,
942 "proxy: client %s (%s) requested Transfer-Encoding "
943 "chunked body with Content-Length (C-L ignored)",
944 c->remote_ip, c->remote_host ? c->remote_host: "");
945 apr_table_unset(r->headers_in, "Content-Length");
947 origin->keepalive = AP_CONN_CLOSE;
951 /* Prefetch MAX_MEM_SPOOL bytes
953 * This helps us avoid any election of C-L v.s. T-E
954 * request bodies, since we are willing to keep in
955 * memory this much data, in any case. This gives
956 * us an instant C-L election if the body is of some
959 temp_brigade = apr_brigade_create(p, bucket_alloc);
961 status = ap_get_brigade(r->input_filters, temp_brigade,
962 AP_MODE_READBYTES, APR_BLOCK_READ,
963 MAX_MEM_SPOOL - bytes_read);
964 if (status != APR_SUCCESS) {
965 ap_log_error(APLOG_MARK, APLOG_ERR, status, r->server,
966 "proxy: prefetch request body failed to %pI (%s)"
968 p_conn->addr, p_conn->hostname ? p_conn->hostname: "",
969 c->remote_ip, c->remote_host ? c->remote_host: "");
970 return HTTP_BAD_REQUEST;
973 apr_brigade_length(temp_brigade, 1, &bytes);
977 * Save temp_brigade in input_brigade. (At least) in the SSL case
978 * temp_brigade contains transient buckets whose data would get
979 * overwritten during the next call of ap_get_brigade in the loop.
980 * ap_save_brigade ensures these buckets to be set aside.
981 * Calling ap_save_brigade with NULL as filter is OK, because
982 * input_brigade already has been created and does not need to get
983 * created by ap_save_brigade.
985 status = ap_save_brigade(NULL, &input_brigade, &temp_brigade, p);
986 if (status != APR_SUCCESS) {
987 ap_log_error(APLOG_MARK, APLOG_ERR, status, r->server,
988 "proxy: processing prefetched request body failed"
989 " to %pI (%s) from %s (%s)",
990 p_conn->addr, p_conn->hostname ? p_conn->hostname: "",
991 c->remote_ip, c->remote_host ? c->remote_host: "");
992 return HTTP_INTERNAL_SERVER_ERROR;
995 /* Ensure we don't hit a wall where we have a buffer too small
996 * for ap_get_brigade's filters to fetch us another bucket,
997 * surrender once we hit 80 bytes less than MAX_MEM_SPOOL
998 * (an arbitrary value.)
1000 } while ((bytes_read < MAX_MEM_SPOOL - 80)
1001 && !APR_BUCKET_IS_EOS(APR_BRIGADE_LAST(input_brigade)));
1003 /* Use chunked request body encoding or send a content-length body?
1007 * We have no request body (handled by RB_STREAM_CL)
1009 * We have a request body length <= MAX_MEM_SPOOL
1011 * The administrator has setenv force-proxy-request-1.0
1013 * The client sent a C-L body, and the administrator has
1014 * not setenv proxy-sendchunked or has set setenv proxy-sendcl
1016 * The client sent a T-E body, and the administrator has
1017 * setenv proxy-sendcl, and not setenv proxy-sendchunked
1019 * If both proxy-sendcl and proxy-sendchunked are set, the
1020 * behavior is the same as if neither were set, large bodies
1021 * that can't be read will be forwarded in their original
1022 * form of C-L, or T-E.
1024 * To ensure maximum compatibility, setenv proxy-sendcl
1025 * To reduce server resource use, setenv proxy-sendchunked
1027 * Then address specific servers with conditional setenv
1028 * options to restore the default behavior where desireable.
1030 * We have to compute content length by reading the entire request
1031 * body; if request body is not small, we'll spool the remaining
1032 * input to a temporary file. Chunked is always preferable.
1034 * We can only trust the client-provided C-L if the T-E header
1035 * is absent, and the filters are unchanged (the body won't
1036 * be resized by another content filter).
1038 if (APR_BUCKET_IS_EOS(APR_BRIGADE_LAST(input_brigade))) {
1039 /* The whole thing fit, so our decision is trivial, use
1040 * the filtered bytes read from the client for the request
1041 * body Content-Length.
1043 * If we expected no body, and read no body, do not set
1044 * the Content-Length.
1046 if (old_cl_val || old_te_val || bytes_read) {
1047 old_cl_val = apr_off_t_toa(r->pool, bytes_read);
1049 rb_method = RB_STREAM_CL;
1051 else if (old_te_val) {
1053 || (apr_table_get(r->subprocess_env, "proxy-sendcl")
1054 && !apr_table_get(r->subprocess_env, "proxy-sendchunks")
1055 && !apr_table_get(r->subprocess_env, "proxy-sendchunked"))) {
1056 rb_method = RB_SPOOL_CL;
1059 rb_method = RB_STREAM_CHUNKED;
1062 else if (old_cl_val) {
1063 if (r->input_filters == r->proto_input_filters) {
1064 rb_method = RB_STREAM_CL;
1067 && (apr_table_get(r->subprocess_env, "proxy-sendchunks")
1068 || apr_table_get(r->subprocess_env, "proxy-sendchunked"))
1069 && !apr_table_get(r->subprocess_env, "proxy-sendcl")) {
1070 rb_method = RB_STREAM_CHUNKED;
1073 rb_method = RB_SPOOL_CL;
1077 /* This is an appropriate default; very efficient for no-body
1078 * requests, and has the behavior that it will not add any C-L
1079 * when the old_cl_val is NULL.
1081 rb_method = RB_SPOOL_CL;
1084 /* Yes I hate gotos. This is the subrequest shortcut */
1087 * Handle Connection: header if we do HTTP/1.1 request:
1088 * If we plan to close the backend connection sent Connection: close
1089 * otherwise sent Connection: Keep-Alive.
1092 if (p_conn->close) {
1093 buf = apr_pstrdup(p, "Connection: close" CRLF);
1096 buf = apr_pstrdup(p, "Connection: Keep-Alive" CRLF);
1098 ap_xlate_proto_to_ascii(buf, strlen(buf));
1099 e = apr_bucket_pool_create(buf, strlen(buf), p, c->bucket_alloc);
1100 APR_BRIGADE_INSERT_TAIL(header_brigade, e);
1103 /* send the request body, if any. */
1105 case RB_STREAM_CHUNKED:
1106 rv = stream_reqbody_chunked(p, r, p_conn, origin, header_brigade,
1110 rv = stream_reqbody_cl(p, r, p_conn, origin, header_brigade,
1111 input_brigade, old_cl_val);
1114 rv = spool_reqbody_cl(p, r, p_conn, origin, header_brigade,
1115 input_brigade, (old_cl_val != NULL)
1116 || (old_te_val != NULL)
1117 || (bytes_read > 0));
1120 /* shouldn't be possible */
1121 rv = HTTP_INTERNAL_SERVER_ERROR ;
1126 /* apr_errno value has been logged in lower level method */
1127 ap_log_error(APLOG_MARK, APLOG_ERR, 0, r->server,
1128 "proxy: pass request body failed to %pI (%s)"
1131 p_conn->hostname ? p_conn->hostname: "",
1133 c->remote_host ? c->remote_host: "");
1140 static void process_proxy_header(request_rec *r, proxy_dir_conf *c,
1141 const char *key, const char *value)
1143 static const char *date_hdrs[]
1144 = { "Date", "Expires", "Last-Modified", NULL };
1145 static const struct {
1147 ap_proxy_header_reverse_map_fn func;
1148 } transform_hdrs[] = {
1149 { "Location", ap_proxy_location_reverse_map },
1150 { "Content-Location", ap_proxy_location_reverse_map },
1151 { "URI", ap_proxy_location_reverse_map },
1152 { "Destination", ap_proxy_location_reverse_map },
1153 { "Set-Cookie", ap_proxy_cookie_reverse_map },
1157 for (i = 0; date_hdrs[i]; ++i) {
1158 if (!strcasecmp(date_hdrs[i], key)) {
1159 apr_table_add(r->headers_out, key,
1160 ap_proxy_date_canon(r->pool, value));
1164 for (i = 0; transform_hdrs[i].name; ++i) {
1165 if (!strcasecmp(transform_hdrs[i].name, key)) {
1166 apr_table_add(r->headers_out, key,
1167 (*transform_hdrs[i].func)(r, c, value));
1171 apr_table_add(r->headers_out, key, value);
1176 * Note: pread_len is the length of the response that we've mistakenly
1177 * read (assuming that we don't consider that an error via
1178 * ProxyBadHeader StartBody). This depends on buffer actually being
1179 * local storage to the calling code in order for pread_len to make
1180 * any sense at all, since we depend on buffer still containing
1181 * what was read by ap_getline() upon return.
1183 static void ap_proxy_read_headers(request_rec *r, request_rec *rr,
1184 char *buffer, int size,
1185 conn_rec *c, int *pread_len)
1189 char field[MAX_STRING_LEN];
1190 int saw_headers = 0;
1191 void *sconf = r->server->module_config;
1192 proxy_server_conf *psc;
1193 proxy_dir_conf *dconf;
1195 dconf = ap_get_module_config(r->per_dir_config, &proxy_module);
1196 psc = (proxy_server_conf *) ap_get_module_config(sconf, &proxy_module);
1198 r->headers_out = apr_table_make(r->pool, 20);
1202 * Read header lines until we get the empty separator line, a read error,
1203 * the connection closes (EOF), or we timeout.
1205 while ((len = ap_getline(buffer, size, rr, 1)) > 0) {
1207 if (!(value = strchr(buffer, ':'))) { /* Find the colon separator */
1209 /* We may encounter invalid headers, usually from buggy
1210 * MS IIS servers, so we need to determine just how to handle
1211 * them. We can either ignore them, assume that they mark the
1212 * start-of-body (eg: a missing CRLF) or (the default) mark
1213 * the headers as totally bogus and return a 500. The sole
1214 * exception is an extra "HTTP/1.0 200, OK" line sprinkled
1215 * in between the usual MIME headers, which is a favorite
1218 /* XXX: The mask check is buggy if we ever see an HTTP/1.10 */
1220 if (!apr_date_checkmask(buffer, "HTTP/#.# ###*")) {
1221 if (psc->badopt == bad_error) {
1222 /* Nope, it wasn't even an extra HTTP header. Give up. */
1223 r->headers_out = NULL;
1226 else if (psc->badopt == bad_body) {
1227 /* if we've already started loading headers_out, then
1228 * return what we've accumulated so far, in the hopes
1229 * that they are useful; also note that we likely pre-read
1230 * the first line of the response.
1233 ap_log_error(APLOG_MARK, APLOG_WARNING, 0, r->server,
1234 "proxy: Starting body due to bogus non-header in headers "
1235 "returned by %s (%s)", r->uri, r->method);
1239 ap_log_error(APLOG_MARK, APLOG_WARNING, 0, r->server,
1240 "proxy: No HTTP headers "
1241 "returned by %s (%s)", r->uri, r->method);
1246 /* this is the psc->badopt == bad_ignore case */
1247 ap_log_error(APLOG_MARK, APLOG_WARNING, 0, r->server,
1248 "proxy: Ignoring bogus HTTP header "
1249 "returned by %s (%s)", r->uri, r->method);
1255 /* XXX: RFC2068 defines only SP and HT as whitespace, this test is
1256 * wrong... and so are many others probably.
1258 while (apr_isspace(*value))
1259 ++value; /* Skip to start of value */
1261 /* should strip trailing whitespace as well */
1262 for (end = &value[strlen(value)-1]; end > value && apr_isspace(*end); --
1266 /* make sure we add so as not to destroy duplicated headers
1267 * Modify headers requiring canonicalisation and/or affected
1268 * by ProxyPassReverse and family with process_proxy_header
1270 process_proxy_header(r, dconf, buffer, value) ;
1273 /* the header was too long; at the least we should skip extra data */
1274 if (len >= size - 1) {
1275 while ((len = ap_getline(field, MAX_STRING_LEN, rr, 1))
1276 >= MAX_STRING_LEN - 1) {
1277 /* soak up the extra data */
1279 if (len == 0) /* time to exit the larger loop as well */
1287 static int addit_dammit(void *v, const char *key, const char *val)
1289 apr_table_addn(v, key, val);
1294 apr_status_t ap_proxygetline(apr_bucket_brigade *bb, char *s, int n, request_rec *r,
1295 int fold, int *writen)
1301 rv = ap_rgetline(&tmp_s, n, &len, r, fold, bb);
1302 apr_brigade_cleanup(bb);
1304 if (rv == APR_SUCCESS) {
1305 *writen = (int) len;
1306 } else if (rv == APR_ENOSPC) {
1316 apr_status_t ap_proxy_http_process_response(apr_pool_t * p, request_rec *r,
1317 proxy_conn_rec *backend,
1319 proxy_server_conf *conf,
1320 char *server_portstr) {
1321 conn_rec *c = r->connection;
1322 char buffer[HUGE_STRING_LEN];
1327 apr_bucket_brigade *bb, *tmp_bb;
1328 int len, backasswards;
1329 int interim_response; /* non-zero whilst interim 1xx responses
1330 * are being read. */
1332 apr_table_t *save_table;
1333 int backend_broke = 0;
1334 static const char *hop_by_hop_hdrs[] =
1335 {"Keep-Alive", "Proxy-Authenticate", "TE", "Trailer", "Upgrade", NULL};
1337 const char *te = NULL;
1339 bb = apr_brigade_create(p, c->bucket_alloc);
1341 /* Get response from the remote server, and pass it up the
1345 rp = ap_proxy_make_fake_req(origin, r);
1346 /* In case anyone needs to know, this is a fake request that is really a
1349 rp->proxyreq = PROXYREQ_RESPONSE;
1350 tmp_bb = apr_brigade_create(p, c->bucket_alloc);
1354 apr_brigade_cleanup(bb);
1356 rc = ap_proxygetline(tmp_bb, buffer, sizeof(buffer), rp, 0, &len);
1358 /* handle one potential stray CRLF */
1359 rc = ap_proxygetline(tmp_bb, buffer, sizeof(buffer), rp, 0, &len);
1362 ap_log_rerror(APLOG_MARK, APLOG_ERR, rc, r,
1363 "proxy: error reading status line from remote "
1364 "server %s", backend->hostname);
1366 * If we are a reverse proxy request shutdown the connection
1367 * WITHOUT ANY response to trigger a retry by the client
1368 * if allowed (as for idempotent requests).
1369 * BUT currently we should not do this if the request is the
1370 * first request on a keepalive connection as browsers like
1371 * seamonkey only display an empty page in this case and do
1374 if (r->proxyreq == PROXYREQ_REVERSE && c->keepalives) {
1377 ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r,
1378 "proxy: Closing connection to client because"
1379 " reading from backend server %s failed. Number"
1380 " of keepalives %i", backend->hostname,
1382 ap_proxy_backend_broke(r, bb);
1384 * Add an EOC bucket to signal the ap_http_header_filter
1385 * that it should get out of our way, BUT ensure that the
1386 * EOC bucket is inserted BEFORE an EOS bucket in bb as
1387 * some resource filters like mod_deflate pass everything
1388 * up to the EOS down the chain immediately and sent the
1389 * remainder of the brigade later (or even never). But in
1390 * this case the ap_http_header_filter does not get out of
1391 * our way soon enough.
1393 e = ap_bucket_eoc_create(c->bucket_alloc);
1394 eos = APR_BRIGADE_LAST(bb);
1395 while ((APR_BRIGADE_SENTINEL(bb) != eos)
1396 && !APR_BUCKET_IS_EOS(eos)) {
1397 eos = APR_BUCKET_PREV(eos);
1399 if (eos == APR_BRIGADE_SENTINEL(bb)) {
1400 APR_BRIGADE_INSERT_TAIL(bb, e);
1403 APR_BUCKET_INSERT_BEFORE(eos, e);
1405 ap_pass_brigade(r->output_filters, bb);
1406 /* Need to return OK to avoid sending an error message */
1409 else if (!c->keepalives) {
1410 ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r,
1411 "proxy: NOT Closing connection to client"
1412 " although reading from backend server %s"
1413 " failed.", backend->hostname);
1415 return ap_proxyerror(r, HTTP_BAD_GATEWAY,
1416 "Error reading from remote server");
1418 /* XXX: Is this a real headers length send from remote? */
1419 backend->worker->s->read += len;
1421 /* Is it an HTTP/1 response?
1422 * This is buggy if we ever see an HTTP/1.10
1424 if (apr_date_checkmask(buffer, "HTTP/#.# ###*")) {
1427 if (2 != sscanf(buffer, "HTTP/%u.%u", &major, &minor)) {
1431 /* If not an HTTP/1 message or
1432 * if the status line was > 8192 bytes
1434 else if ((buffer[5] != '1') || (len >= sizeof(buffer)-1)) {
1435 return ap_proxyerror(r, HTTP_BAD_GATEWAY,
1436 apr_pstrcat(p, "Corrupt status line returned by remote "
1437 "server: ", buffer, NULL));
1441 keepchar = buffer[12];
1443 r->status = atoi(&buffer[9]);
1445 if (keepchar != '\0') {
1446 buffer[12] = keepchar;
1448 /* 2616 requires the space in Status-Line; the origin
1449 * server may have sent one but ap_rgetline_core will
1450 * have stripped it. */
1454 r->status_line = apr_pstrdup(p, &buffer[9]);
1457 /* read the headers. */
1458 /* N.B. for HTTP/1.0 clients, we have to fold line-wrapped headers*/
1459 /* Also, take care with headers with multiple occurences. */
1461 /* First, tuck away all already existing cookies */
1462 save_table = apr_table_make(r->pool, 2);
1463 apr_table_do(addit_dammit, save_table, r->headers_out,
1464 "Set-Cookie", NULL);
1466 /* shove the headers direct into r->headers_out */
1467 ap_proxy_read_headers(r, rp, buffer, sizeof(buffer), origin,
1470 if (r->headers_out == NULL) {
1471 ap_log_error(APLOG_MARK, APLOG_WARNING, 0,
1472 r->server, "proxy: bad HTTP/%d.%d header "
1473 "returned by %s (%s)", major, minor, r->uri,
1475 backend->close += 1;
1477 * ap_send_error relies on a headers_out to be present. we
1478 * are in a bad position here.. so force everything we send out
1479 * to have nothing to do with the incoming packet
1481 r->headers_out = apr_table_make(r->pool,1);
1482 r->status = HTTP_BAD_GATEWAY;
1483 r->status_line = "bad gateway";
1487 /* Now, add in the just read cookies */
1488 apr_table_do(addit_dammit, save_table, r->headers_out,
1489 "Set-Cookie", NULL);
1491 /* and now load 'em all in */
1492 if (!apr_is_empty_table(save_table)) {
1493 apr_table_unset(r->headers_out, "Set-Cookie");
1494 r->headers_out = apr_table_overlay(r->pool,
1499 /* can't have both Content-Length and Transfer-Encoding */
1500 if (apr_table_get(r->headers_out, "Transfer-Encoding")
1501 && apr_table_get(r->headers_out, "Content-Length")) {
1503 * 2616 section 4.4, point 3: "if both Transfer-Encoding
1504 * and Content-Length are received, the latter MUST be
1507 * To help mitigate HTTP Splitting, unset Content-Length
1508 * and shut down the backend server connection
1509 * XXX: We aught to treat such a response as uncachable
1511 apr_table_unset(r->headers_out, "Content-Length");
1512 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
1513 "proxy: server %s returned Transfer-Encoding"
1514 " and Content-Length", backend->hostname);
1515 backend->close += 1;
1519 * Save a possible Transfer-Encoding header as we need it later for
1520 * ap_http_filter to know where to end.
1522 te = apr_table_get(r->headers_out, "Transfer-Encoding");
1523 /* strip connection listed hop-by-hop headers from response */
1524 backend->close += ap_proxy_liststr(apr_table_get(r->headers_out,
1527 ap_proxy_clear_connection(p, r->headers_out);
1528 if ((buf = apr_table_get(r->headers_out, "Content-Type"))) {
1529 ap_set_content_type(r, apr_pstrdup(p, buf));
1531 ap_proxy_pre_http_request(origin,rp);
1533 /* Clear hop-by-hop headers */
1534 for (i=0; hop_by_hop_hdrs[i]; ++i) {
1535 apr_table_unset(r->headers_out, hop_by_hop_hdrs[i]);
1537 /* Delete warnings with wrong date */
1538 r->headers_out = ap_proxy_clean_warnings(p, r->headers_out);
1540 /* handle Via header in response */
1541 if (conf->viaopt != via_off && conf->viaopt != via_block) {
1542 const char *server_name = ap_get_server_name(r);
1543 /* If USE_CANONICAL_NAME_OFF was configured for the proxy virtual host,
1544 * then the server name returned by ap_get_server_name() is the
1545 * origin server name (which does make too much sense with Via: headers)
1546 * so we use the proxy vhost's name instead.
1548 if (server_name == r->hostname)
1549 server_name = r->server->server_hostname;
1550 /* create a "Via:" response header entry and merge it */
1551 apr_table_addn(r->headers_out, "Via",
1552 (conf->viaopt == via_full)
1553 ? apr_psprintf(p, "%d.%d %s%s (%s)",
1554 HTTP_VERSION_MAJOR(r->proto_num),
1555 HTTP_VERSION_MINOR(r->proto_num),
1558 AP_SERVER_BASEVERSION)
1559 : apr_psprintf(p, "%d.%d %s%s",
1560 HTTP_VERSION_MAJOR(r->proto_num),
1561 HTTP_VERSION_MINOR(r->proto_num),
1567 /* cancel keepalive if HTTP/1.0 or less */
1568 if ((major < 1) || (minor < 1)) {
1569 backend->close += 1;
1570 origin->keepalive = AP_CONN_CLOSE;
1573 /* an http/0.9 response */
1576 r->status_line = "200 OK";
1577 backend->close += 1;
1580 interim_response = ap_is_HTTP_INFO(r->status);
1581 if (interim_response) {
1582 /* RFC2616 tells us to forward this.
1584 * OTOH, an interim response here may mean the backend
1585 * is playing sillybuggers. The Client didn't ask for
1586 * it within the defined HTTP/1.1 mechanisms, and if
1587 * it's an extension, it may also be unsupported by us.
1589 * There's also the possibility that changing existing
1590 * behaviour here might break something.
1592 * So let's make it configurable.
1594 const char *policy = apr_table_get(r->subprocess_env,
1595 "proxy-interim-response");
1596 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, NULL,
1597 "proxy: HTTP: received interim %d response",
1599 if (!policy || !strcasecmp(policy, "RFC")) {
1600 ap_send_interim_response(r, 1);
1602 /* FIXME: refine this to be able to specify per-response-status
1603 * policies and maybe also add option to bail out with 502
1605 else if (strcasecmp(policy, "Suppress")) {
1606 ap_log_error(APLOG_MARK, APLOG_WARNING, 0, NULL,
1607 "undefined proxy interim response policy");
1610 /* Moved the fixups of Date headers and those affected by
1611 * ProxyPassReverse/etc from here to ap_proxy_read_headers
1614 if ((r->status == 401) && (conf->error_override)) {
1616 const char *wa = "WWW-Authenticate";
1617 if ((buf = apr_table_get(r->headers_out, wa))) {
1618 apr_table_set(r->err_headers_out, wa, buf);
1620 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
1621 "proxy: origin server sent 401 without WWW-Authenticate header");
1627 * Is it an HTTP/0.9 response or did we maybe preread the 1st line of
1628 * the response? If so, load the extra data. These are 2 mutually
1629 * exclusive possibilities, that just happen to require very
1632 if (backasswards || pread_len) {
1633 apr_ssize_t cntr = (apr_ssize_t)pread_len;
1636 * At this point in response processing of a 0.9 response,
1637 * we don't know yet whether data is binary or not.
1638 * mod_charset_lite will get control later on, so it cannot
1639 * decide on the conversion of this buffer full of data.
1640 * However, chances are that we are not really talking to an
1641 * HTTP/0.9 server, but to some different protocol, therefore
1642 * the best guess IMHO is to always treat the buffer as "text/x":
1644 ap_xlate_proto_to_ascii(buffer, len);
1645 cntr = (apr_ssize_t)len;
1647 e = apr_bucket_heap_create(buffer, cntr, NULL, c->bucket_alloc);
1648 APR_BRIGADE_INSERT_TAIL(bb, e);
1651 /* send body - but only if a body is expected */
1652 if ((!r->header_only) && /* not HEAD request */
1653 !interim_response && /* not any 1xx response */
1654 (r->status != HTTP_NO_CONTENT) && /* not 204 */
1655 (r->status != HTTP_NOT_MODIFIED)) { /* not 304 */
1657 /* We need to copy the output headers and treat them as input
1658 * headers as well. BUT, we need to do this before we remove
1659 * TE, so that they are preserved accordingly for
1660 * ap_http_filter to know where to end.
1662 rp->headers_in = apr_table_copy(r->pool, r->headers_out);
1664 * Restore Transfer-Encoding header from response if we saved
1665 * one before and there is none left. We need it for the
1666 * ap_http_filter. See above.
1668 if (te && !apr_table_get(rp->headers_in, "Transfer-Encoding")) {
1669 apr_table_add(rp->headers_in, "Transfer-Encoding", te);
1672 apr_table_unset(r->headers_out,"Transfer-Encoding");
1674 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
1675 "proxy: start body send");
1678 * if we are overriding the errors, we can't put the content
1679 * of the page into the brigade
1681 if (!conf->error_override || !ap_is_HTTP_ERROR(r->status)) {
1682 /* read the body, pass it to the output filters */
1683 apr_read_type_e mode = APR_NONBLOCK_READ;
1687 apr_off_t readbytes;
1690 rv = ap_get_brigade(rp->input_filters, bb,
1691 AP_MODE_READBYTES, mode,
1692 conf->io_buffer_size);
1694 /* ap_get_brigade will return success with an empty brigade
1695 * for a non-blocking read which would block: */
1696 if (APR_STATUS_IS_EAGAIN(rv)
1697 || (rv == APR_SUCCESS && APR_BRIGADE_EMPTY(bb))) {
1698 /* flush to the client and switch to blocking mode */
1699 e = apr_bucket_flush_create(c->bucket_alloc);
1700 APR_BRIGADE_INSERT_TAIL(bb, e);
1701 if (ap_pass_brigade(r->output_filters, bb)
1706 apr_brigade_cleanup(bb);
1707 mode = APR_BLOCK_READ;
1710 else if (rv == APR_EOF) {
1713 else if (rv != APR_SUCCESS) {
1714 /* In this case, we are in real trouble because
1715 * our backend bailed on us. Pass along a 502 error
1718 ap_log_cerror(APLOG_MARK, APLOG_ERR, rv, c,
1719 "proxy: error reading response");
1720 ap_proxy_backend_broke(r, bb);
1721 ap_pass_brigade(r->output_filters, bb);
1726 /* next time try a non-blocking read */
1727 mode = APR_NONBLOCK_READ;
1729 apr_brigade_length(bb, 0, &readbytes);
1730 backend->worker->s->read += readbytes;
1733 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0,
1734 r->server, "proxy (PID %d): readbytes: %#x",
1735 getpid(), readbytes);
1739 if (APR_BRIGADE_EMPTY(bb)) {
1740 apr_brigade_cleanup(bb);
1744 /* found the last brigade? */
1745 if (APR_BUCKET_IS_EOS(APR_BRIGADE_LAST(bb))) {
1746 /* signal that we must leave */
1750 /* try send what we read */
1751 if (ap_pass_brigade(r->output_filters, bb) != APR_SUCCESS
1753 /* Ack! Phbtt! Die! User aborted! */
1754 backend->close = 1; /* this causes socket close below */
1758 /* make sure we always clean up after ourselves */
1759 apr_brigade_cleanup(bb);
1763 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
1764 "proxy: end body send");
1766 else if (!interim_response) {
1767 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
1768 "proxy: header only");
1770 /* Pass EOS bucket down the filter chain. */
1771 e = apr_bucket_eos_create(c->bucket_alloc);
1772 APR_BRIGADE_INSERT_TAIL(bb, e);
1773 if (ap_pass_brigade(r->output_filters, bb) != APR_SUCCESS
1775 /* Ack! Phbtt! Die! User aborted! */
1776 backend->close = 1; /* this causes socket close below */
1779 apr_brigade_cleanup(bb);
1781 } while (interim_response);
1783 /* If our connection with the client is to be aborted, return DONE. */
1784 if (c->aborted || backend_broke) {
1788 if (conf->error_override) {
1789 /* the code above this checks for 'OK' which is what the hook expects */
1790 if (!ap_is_HTTP_ERROR(r->status))
1793 /* clear r->status for override error, otherwise ErrorDocument
1794 * thinks that this is a recursive error, and doesn't find the
1797 int status = r->status;
1798 r->status = HTTP_OK;
1799 /* Discard body, if one is expected */
1800 if (!r->header_only && /* not HEAD request */
1801 (status != HTTP_NO_CONTENT) && /* not 204 */
1802 (status != HTTP_NOT_MODIFIED)) { /* not 304 */
1803 ap_discard_request_body(rp);
1812 apr_status_t ap_proxy_http_cleanup(const char *scheme, request_rec *r,
1813 proxy_conn_rec *backend)
1815 ap_proxy_release_connection(scheme, backend, r->server);
1820 * This handles http:// URLs, and other URLs using a remote proxy over http
1821 * If proxyhost is NULL, then contact the server directly, otherwise
1823 * Note that if a proxy is used, then URLs other than http: can be accessed,
1824 * also, if we have trouble which is clearly specific to the proxy, then
1825 * we return DECLINED so that we can try another proxy. (Or the direct
1828 static int proxy_http_handler(request_rec *r, proxy_worker *worker,
1829 proxy_server_conf *conf,
1830 char *url, const char *proxyname,
1831 apr_port_t proxyport)
1834 char server_portstr[32];
1836 const char *proxy_function;
1838 proxy_conn_rec *backend = NULL;
1841 /* Note: Memory pool allocation.
1842 * A downstream keepalive connection is always connected to the existence
1843 * (or not) of an upstream keepalive connection. If this is not done then
1844 * load balancing against multiple backend servers breaks (one backend
1845 * server ends up taking 100% of the load), and the risk is run of
1846 * downstream keepalive connections being kept open unnecessarily. This
1847 * keeps webservers busy and ties up resources.
1849 * As a result, we allocate all sockets out of the upstream connection
1850 * pool, and when we want to reuse a socket, we check first whether the
1851 * connection ID of the current upstream connection is the same as that
1852 * of the connection when the socket was opened.
1854 apr_pool_t *p = r->connection->pool;
1855 conn_rec *c = r->connection;
1856 apr_uri_t *uri = apr_palloc(r->connection->pool, sizeof(*uri));
1858 /* find the scheme */
1859 u = strchr(url, ':');
1860 if (u == NULL || u[1] != '/' || u[2] != '/' || u[3] == '\0')
1863 return HTTP_BAD_REQUEST;
1864 scheme = apr_pstrndup(c->pool, url, u - url);
1865 /* scheme is lowercase */
1866 ap_str_tolower(scheme);
1868 if (strcmp(scheme, "https") == 0) {
1869 if (!ap_proxy_ssl_enable(NULL)) {
1870 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
1871 "proxy: HTTPS: declining URL %s"
1872 " (mod_ssl not configured?)", url);
1876 proxy_function = "HTTPS";
1878 else if (!(strcmp(scheme, "http") == 0 || (strcmp(scheme, "ftp") == 0 && proxyname))) {
1879 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
1880 "proxy: HTTP: declining URL %s", url);
1881 return DECLINED; /* only interested in HTTP, or FTP via proxy */
1885 proxy_function = "HTTP";
1887 proxy_function = "FTP";
1889 ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, r->server,
1890 "proxy: HTTP: serving URL %s", url);
1893 /* create space for state information */
1894 if ((status = ap_proxy_acquire_connection(proxy_function, &backend,
1895 worker, r->server)) != OK)
1899 backend->is_ssl = is_ssl;
1902 ap_proxy_ssl_connection_cleanup(backend, r);
1905 /* Step One: Determine Who To Connect To */
1906 if ((status = ap_proxy_determine_connection(p, r, conf, worker, backend,
1907 uri, &url, proxyname,
1908 proxyport, server_portstr,
1909 sizeof(server_portstr))) != OK)
1912 /* Step Two: Make the Connection */
1913 if (ap_proxy_connect_backend(proxy_function, backend, worker, r->server)) {
1914 if (r->proxyreq == PROXYREQ_PROXY)
1915 status = HTTP_NOT_FOUND;
1917 status = HTTP_SERVICE_UNAVAILABLE;
1921 /* Step Three: Create conn_rec */
1922 if (!backend->connection) {
1923 if ((status = ap_proxy_connection_create(proxy_function, backend,
1924 c, r->server)) != OK)
1928 /* Step Four: Send the Request */
1929 if ((status = ap_proxy_http_request(p, r, backend, backend->connection,
1930 conf, uri, url, server_portstr)) != OK)
1933 /* Step Five: Receive the Response */
1934 if ((status = ap_proxy_http_process_response(p, r, backend,
1935 backend->connection,
1936 conf, server_portstr)) != OK)
1939 /* Step Six: Clean Up */
1945 ap_proxy_http_cleanup(proxy_function, r, backend);
1949 static apr_status_t warn_rx_free(void *p)
1951 ap_pregfree((apr_pool_t*)p, warn_rx);
1954 static void ap_proxy_http_register_hook(apr_pool_t *p)
1956 proxy_hook_scheme_handler(proxy_http_handler, NULL, NULL, APR_HOOK_FIRST);
1957 proxy_hook_canon_handler(proxy_http_canon, NULL, NULL, APR_HOOK_FIRST);
1958 warn_rx = ap_pregcomp(p, "[0-9]{3}[ \t]+[^ \t]+[ \t]+\"[^\"]*\"([ \t]+\"([^\"]+)\")?", 0);
1959 apr_pool_cleanup_register(p, p, warn_rx_free, apr_pool_cleanup_null);
1962 module AP_MODULE_DECLARE_DATA proxy_http_module = {
1963 STANDARD20_MODULE_STUFF,
1964 NULL, /* create per-directory config structure */
1965 NULL, /* merge per-directory config structures */
1966 NULL, /* create per-server config structure */
1967 NULL, /* merge per-server config structures */
1968 NULL, /* command apr_table_t */
1969 ap_proxy_http_register_hook/* register hooks */