diff options
Diffstat (limited to '')
| -rw-r--r-- | src/http_ana.c | 5277 |
1 files changed, 5277 insertions, 0 deletions
diff --git a/src/http_ana.c b/src/http_ana.c new file mode 100644 index 0000000..273fe16 --- /dev/null +++ b/src/http_ana.c @@ -0,0 +1,5277 @@ +/* + * HTTP protocol analyzer + * + * Copyright (C) 2018 HAProxy Technologies, Christopher Faulet <cfaulet@haproxy.com> + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + * + */ + +#include <haproxy/acl.h> +#include <haproxy/action-t.h> +#include <haproxy/api.h> +#include <haproxy/applet.h> +#include <haproxy/backend.h> +#include <haproxy/base64.h> +#include <haproxy/capture-t.h> +#include <haproxy/cfgparse.h> +#include <haproxy/channel.h> +#include <haproxy/check.h> +#include <haproxy/connection.h> +#include <haproxy/errors.h> +#include <haproxy/filters.h> +#include <haproxy/http.h> +#include <haproxy/http_ana.h> +#include <haproxy/http_htx.h> +#include <haproxy/htx.h> +#include <haproxy/log.h> +#include <haproxy/net_helper.h> +#include <haproxy/proxy.h> +#include <haproxy/regex.h> +#include <haproxy/sc_strm.h> +#include <haproxy/server-t.h> +#include <haproxy/stats.h> +#include <haproxy/stconn.h> +#include <haproxy/stream.h> +#include <haproxy/trace.h> +#include <haproxy/uri_auth-t.h> +#include <haproxy/vars.h> + + +#define TRACE_SOURCE &trace_strm + +extern const char *stat_status_codes[]; + +struct pool_head *pool_head_requri __read_mostly = NULL; +struct pool_head *pool_head_capture __read_mostly = NULL; + + +static void http_end_request(struct stream *s); +static void http_end_response(struct stream *s); + +static void http_capture_headers(struct htx *htx, char **cap, struct cap_hdr *cap_hdr); +static int http_del_hdr_value(char *start, char *end, char **from, char *next); +static size_t http_fmt_req_line(const struct htx_sl *sl, char *str, size_t len); +static void http_debug_stline(const char *dir, struct stream *s, const struct htx_sl *sl); +static void http_debug_hdr(const char *dir, struct stream *s, const struct ist n, const struct ist v); + +static enum rule_result http_req_get_intercept_rule(struct proxy *px, struct list *def_rules, struct list *rules, struct stream *s); +static enum rule_result http_res_get_intercept_rule(struct proxy *px, struct list *def_rules, struct list *rules, struct stream *s); +static enum rule_result http_req_restrict_header_names(struct stream *s, struct htx *htx, struct proxy *px); + +static void http_manage_client_side_cookies(struct stream *s, struct channel *req); +static void http_manage_server_side_cookies(struct stream *s, struct channel *res); + +static int http_stats_check_uri(struct stream *s, struct http_txn *txn, struct proxy *backend); +static int http_handle_stats(struct stream *s, struct channel *req); + +static int http_handle_expect_hdr(struct stream *s, struct htx *htx, struct http_msg *msg); +static int http_reply_100_continue(struct stream *s); + +/* This stream analyser waits for a complete HTTP request. It returns 1 if the + * processing can continue on next analysers, or zero if it either needs more + * data or wants to immediately abort the request (eg: timeout, error, ...). It + * is tied to AN_REQ_WAIT_HTTP and may may remove itself from s->req.analysers + * when it has nothing left to do, and may remove any analyser when it wants to + * abort. + */ +int http_wait_for_request(struct stream *s, struct channel *req, int an_bit) +{ + + /* + * We will analyze a complete HTTP request to check the its syntax. + * + * Once the start line and all headers are received, we may perform a + * capture of the error (if any), and we will set a few fields. We also + * check for monitor-uri, logging and finally headers capture. + */ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct http_msg *msg = &txn->req; + struct htx *htx; + struct htx_sl *sl; + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn, msg); + + if (unlikely(!IS_HTX_STRM(s))) { + /* It is only possible when a TCP stream is upgrade to HTTP. + * There is a transition period during which there is no + * data. The stream is still in raw mode and SF_IGNORE flag is + * still set. When this happens, the new mux is responsible to + * handle all errors. Thus we may leave immediately. + */ + BUG_ON(!(s->flags & SF_IGNORE) || !c_empty(&s->req)); + + /* Don't connect for now */ + channel_dont_connect(req); + + /* A SHUTR at this stage means we are performing a "destructive" + * HTTP upgrade (TCP>H2). In this case, we can leave. + */ + if (req->flags & CF_SHUTR) { + s->logs.logwait = 0; + s->logs.level = 0; + channel_abort(&s->req); + channel_abort(&s->res); + req->analysers &= AN_REQ_FLT_END; + req->analyse_exp = TICK_ETERNITY; + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA, s); + return 1; + } + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA, s); + return 0; + } + + htx = htxbuf(&req->buf); + + /* Parsing errors are caught here */ + if (htx->flags & (HTX_FL_PARSING_ERROR|HTX_FL_PROCESSING_ERROR)) { + stream_inc_http_req_ctr(s); + proxy_inc_fe_req_ctr(sess->listener, sess->fe); + if (htx->flags & HTX_FL_PARSING_ERROR) { + stream_inc_http_err_ctr(s); + goto return_bad_req; + } + else + goto return_int_err; + } + + /* we're speaking HTTP here, so let's speak HTTP to the client */ + s->srv_error = http_return_srv_error; + + msg->msg_state = HTTP_MSG_BODY; + stream_inc_http_req_ctr(s); + proxy_inc_fe_req_ctr(sess->listener, sess->fe); /* one more valid request for this FE */ + + /* kill the pending keep-alive timeout */ + req->analyse_exp = TICK_ETERNITY; + + BUG_ON(htx_get_first_type(htx) != HTX_BLK_REQ_SL); + sl = http_get_stline(htx); + + /* 0: we might have to print this header in debug mode */ + if (unlikely((global.mode & MODE_DEBUG) && + (!(global.mode & MODE_QUIET) || (global.mode & MODE_VERBOSE)))) { + int32_t pos; + + http_debug_stline("clireq", s, sl); + + for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) { + struct htx_blk *blk = htx_get_blk(htx, pos); + enum htx_blk_type type = htx_get_blk_type(blk); + + if (type == HTX_BLK_EOH) + break; + if (type != HTX_BLK_HDR) + continue; + + http_debug_hdr("clihdr", s, + htx_get_blk_name(htx, blk), + htx_get_blk_value(htx, blk)); + } + } + + /* + * 1: identify the method and the version. Also set HTTP flags + */ + txn->meth = sl->info.req.meth; + if (sl->flags & HTX_SL_F_VER_11) + msg->flags |= HTTP_MSGF_VER_11; + msg->flags |= HTTP_MSGF_XFER_LEN; + if (sl->flags & HTX_SL_F_CLEN) + msg->flags |= HTTP_MSGF_CNT_LEN; + else if (sl->flags & HTX_SL_F_CHNK) + msg->flags |= HTTP_MSGF_TE_CHNK; + if (sl->flags & HTX_SL_F_BODYLESS) + msg->flags |= HTTP_MSGF_BODYLESS; + if (sl->flags & HTX_SL_F_CONN_UPG) + msg->flags |= HTTP_MSGF_CONN_UPG; + + /* we can make use of server redirect on GET and HEAD */ + if (txn->meth == HTTP_METH_GET || txn->meth == HTTP_METH_HEAD) + s->flags |= SF_REDIRECTABLE; + else if (txn->meth == HTTP_METH_OTHER && isteqi(htx_sl_req_meth(sl), ist("PRI"))) { + /* PRI is reserved for the HTTP/2 preface */ + goto return_bad_req; + } + + /* + * 2: check if the URI matches the monitor_uri. We have to do this for + * every request which gets in, because the monitor-uri is defined by + * the frontend. If the monitor-uri starts with a '/', the matching is + * done against the request's path. Otherwise, the request's uri is + * used. It is a workaround to let HTTP/2 health-checks work as + * expected. + */ + if (unlikely(isttest(sess->fe->monitor_uri))) { + const struct ist monitor_uri = sess->fe->monitor_uri; + struct http_uri_parser parser = http_uri_parser_init(htx_sl_req_uri(sl)); + + if ((istptr(monitor_uri)[0] == '/' && + isteq(http_parse_path(&parser), monitor_uri)) || + isteq(htx_sl_req_uri(sl), monitor_uri)) { + /* + * We have found the monitor URI + */ + struct acl_cond *cond; + + s->flags |= SF_MONITOR; + _HA_ATOMIC_INC(&sess->fe->fe_counters.intercepted_req); + + /* Check if we want to fail this monitor request or not */ + list_for_each_entry(cond, &sess->fe->mon_fail_cond, list) { + int ret = acl_exec_cond(cond, sess->fe, sess, s, SMP_OPT_DIR_REQ|SMP_OPT_FINAL); + + ret = acl_pass(ret); + if (cond->pol == ACL_COND_UNLESS) + ret = !ret; + + if (ret) { + /* we fail this request, let's return 503 service unavail */ + txn->status = 503; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_LOCAL; /* we don't want a real error here */ + goto return_prx_cond; + } + } + + /* nothing to fail, let's reply normally */ + txn->status = 200; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_LOCAL; /* we don't want a real error here */ + goto return_prx_cond; + } + } + + /* + * 3: Maybe we have to copy the original REQURI for the logs ? + * Note: we cannot log anymore if the request has been + * classified as invalid. + */ + if (unlikely(s->logs.logwait & LW_REQ)) { + /* we have a complete HTTP request that we must log */ + if ((txn->uri = pool_alloc(pool_head_requri)) != NULL) { + size_t len; + + len = http_fmt_req_line(sl, txn->uri, global.tune.requri_len - 1); + txn->uri[len] = 0; + + if (!(s->logs.logwait &= ~(LW_REQ|LW_INIT))) + s->do_log(s); + } else { + ha_alert("HTTP logging : out of memory.\n"); + } + } + + /* if the frontend has "option http-use-proxy-header", we'll check if + * we have what looks like a proxied connection instead of a connection, + * and in this case set the TX_USE_PX_CONN flag to use Proxy-connection. + * Note that this is *not* RFC-compliant, however browsers and proxies + * happen to do that despite being non-standard :-( + * We consider that a request not beginning with either '/' or '*' is + * a proxied connection, which covers both "scheme://location" and + * CONNECT ip:port. + */ + if ((sess->fe->options2 & PR_O2_USE_PXHDR) && + *HTX_SL_REQ_UPTR(sl) != '/' && *HTX_SL_REQ_UPTR(sl) != '*') + txn->flags |= TX_USE_PX_CONN; + + /* 5: we may need to capture headers */ + if (unlikely((s->logs.logwait & LW_REQHDR) && s->req_cap)) + http_capture_headers(htx, s->req_cap, sess->fe->req_cap); + + /* we may have to wait for the request's body */ + if (s->be->options & PR_O_WREQ_BODY) + req->analysers |= AN_REQ_HTTP_BODY; + + /* + * RFC7234#4: + * A cache MUST write through requests with methods + * that are unsafe (Section 4.2.1 of [RFC7231]) to + * the origin server; i.e., a cache is not allowed + * to generate a reply to such a request before + * having forwarded the request and having received + * a corresponding response. + * + * RFC7231#4.2.1: + * Of the request methods defined by this + * specification, the GET, HEAD, OPTIONS, and TRACE + * methods are defined to be safe. + */ + if (likely(txn->meth == HTTP_METH_GET || + txn->meth == HTTP_METH_HEAD || + txn->meth == HTTP_METH_OPTIONS || + txn->meth == HTTP_METH_TRACE)) + txn->flags |= TX_CACHEABLE | TX_CACHE_COOK; + + /* end of job, return OK */ + req->analysers &= ~an_bit; + req->analyse_exp = TICK_ETERNITY; + + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 1; + + return_int_err: + txn->status = 500; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + _HA_ATOMIC_INC(&sess->fe->fe_counters.internal_errors); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->internal_errors); + goto return_prx_cond; + + return_bad_req: + txn->status = 400; + _HA_ATOMIC_INC(&sess->fe->fe_counters.failed_req); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->failed_req); + /* fall through */ + + return_prx_cond: + http_reply_and_close(s, txn->status, http_error_message(s)); + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_R; + + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; +} + + +/* This stream analyser runs all HTTP request processing which is common to + * frontends and backends, which means blocking ACLs, filters, connection-close, + * reqadd, stats and redirects. This is performed for the designated proxy. + * It returns 1 if the processing can continue on next analysers, or zero if it + * either needs more data or wants to immediately abort the request (eg: deny, + * error, ...). + */ +int http_process_req_common(struct stream *s, struct channel *req, int an_bit, struct proxy *px) +{ + struct list *def_rules, *rules; + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct http_msg *msg = &txn->req; + struct htx *htx; + struct redirect_rule *rule; + enum rule_result verdict; + struct connection *conn = objt_conn(sess->origin); + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn, msg); + + htx = htxbuf(&req->buf); + + /* just in case we have some per-backend tracking. Only called the first + * execution of the analyser. */ + if (!s->current_rule && !s->current_rule_list) + stream_inc_be_http_req_ctr(s); + + def_rules = ((px->defpx && (an_bit == AN_REQ_HTTP_PROCESS_FE || px != sess->fe)) ? &px->defpx->http_req_rules : NULL); + rules = &px->http_req_rules; + + /* evaluate http-request rules */ + if ((def_rules && !LIST_ISEMPTY(def_rules)) || !LIST_ISEMPTY(rules)) { + verdict = http_req_get_intercept_rule(px, def_rules, rules, s); + + switch (verdict) { + case HTTP_RULE_RES_YIELD: /* some data miss, call the function later. */ + goto return_prx_yield; + + case HTTP_RULE_RES_CONT: + case HTTP_RULE_RES_STOP: /* nothing to do */ + break; + + case HTTP_RULE_RES_DENY: /* deny or tarpit */ + if (txn->flags & TX_CLTARPIT) + goto tarpit; + goto deny; + + case HTTP_RULE_RES_ABRT: /* abort request, response already sent. Eg: auth */ + goto return_prx_cond; + + case HTTP_RULE_RES_DONE: /* OK, but terminate request processing (eg: redirect) */ + goto done; + + case HTTP_RULE_RES_BADREQ: /* failed with a bad request */ + goto return_bad_req; + + case HTTP_RULE_RES_ERROR: /* failed with a bad request */ + goto return_int_err; + } + } + + if (px->options2 & (PR_O2_RSTRICT_REQ_HDR_NAMES_BLK|PR_O2_RSTRICT_REQ_HDR_NAMES_DEL)) { + verdict = http_req_restrict_header_names(s, htx, px); + if (verdict == HTTP_RULE_RES_DENY) + goto deny; + } + + if (conn && (conn->flags & CO_FL_EARLY_DATA) && + (conn->flags & (CO_FL_EARLY_SSL_HS | CO_FL_SSL_WAIT_HS))) { + struct http_hdr_ctx ctx; + + ctx.blk = NULL; + if (!http_find_header(htx, ist("Early-Data"), &ctx, 0)) { + if (unlikely(!http_add_header(htx, ist("Early-Data"), ist("1")))) + goto return_fail_rewrite; + } + } + + /* OK at this stage, we know that the request was accepted according to + * the http-request rules, we can check for the stats. Note that the + * URI is detected *before* the req* rules in order not to be affected + * by a possible reqrep, while they are processed *after* so that a + * reqdeny can still block them. This clearly needs to change in 1.6! + */ + if (!s->target && http_stats_check_uri(s, txn, px)) { + s->target = &http_stats_applet.obj_type; + if (unlikely(!sc_applet_create(s->scb, objt_applet(s->target)))) { + s->logs.tv_request = now; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_RESOURCE; + goto return_int_err; + } + + /* parse the whole stats request and extract the relevant information */ + http_handle_stats(s, req); + verdict = http_req_get_intercept_rule(px, NULL, &px->uri_auth->http_req_rules, s); + /* not all actions implemented: deny, allow, auth */ + + if (verdict == HTTP_RULE_RES_DENY) /* stats http-request deny */ + goto deny; + + if (verdict == HTTP_RULE_RES_ABRT) /* stats auth / stats http-request auth */ + goto return_prx_cond; + + if (verdict == HTTP_RULE_RES_BADREQ) /* failed with a bad request */ + goto return_bad_req; + + if (verdict == HTTP_RULE_RES_ERROR) /* failed with a bad request */ + goto return_int_err; + } + + /* Proceed with the applets now. */ + if (unlikely(objt_applet(s->target))) { + if (sess->fe == s->be) /* report it if the request was intercepted by the frontend */ + _HA_ATOMIC_INC(&sess->fe->fe_counters.intercepted_req); + + if (http_handle_expect_hdr(s, htx, msg) == -1) + goto return_int_err; + + if (!(s->flags & SF_ERR_MASK)) // this is not really an error but it is + s->flags |= SF_ERR_LOCAL; // to mark that it comes from the proxy + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_R; + + if (HAS_FILTERS(s)) + req->analysers |= AN_REQ_FLT_HTTP_HDRS; + + /* enable the minimally required analyzers to handle keep-alive and compression on the HTTP response */ + req->analysers &= (AN_REQ_HTTP_BODY | AN_REQ_FLT_HTTP_HDRS | AN_REQ_FLT_END); + req->analysers &= ~AN_REQ_FLT_XFER_DATA; + req->analysers |= AN_REQ_HTTP_XFER_BODY; + + req->flags |= CF_SEND_DONTWAIT; + s->flags |= SF_ASSIGNED; + goto done; + } + + /* check whether we have some ACLs set to redirect this request */ + list_for_each_entry(rule, &px->redirect_rules, list) { + if (rule->cond) { + int ret; + + ret = acl_exec_cond(rule->cond, px, sess, s, SMP_OPT_DIR_REQ|SMP_OPT_FINAL); + ret = acl_pass(ret); + if (rule->cond->pol == ACL_COND_UNLESS) + ret = !ret; + if (!ret) + continue; + } + if (!http_apply_redirect_rule(rule, s, txn)) + goto return_int_err; + goto done; + } + + /* POST requests may be accompanied with an "Expect: 100-Continue" header. + * If this happens, then the data will not come immediately, so we must + * send all what we have without waiting. Note that due to the small gain + * in waiting for the body of the request, it's easier to simply put the + * CF_SEND_DONTWAIT flag any time. It's a one-shot flag so it will remove + * itself once used. + */ + req->flags |= CF_SEND_DONTWAIT; + + done: /* done with this analyser, continue with next ones that the calling + * points will have set, if any. + */ + req->analyse_exp = TICK_ETERNITY; + done_without_exp: /* done with this analyser, but don't reset the analyse_exp. */ + req->analysers &= ~an_bit; + s->current_rule = s->current_rule_list = NULL; + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 1; + + tarpit: + /* Allow cookie logging + */ + if (s->be->cookie_name || sess->fe->capture_name) + http_manage_client_side_cookies(s, req); + + /* When a connection is tarpitted, we use the tarpit timeout, + * which may be the same as the connect timeout if unspecified. + * If unset, then set it to zero because we really want it to + * eventually expire. We build the tarpit as an analyser. + */ + channel_htx_erase(&s->req, htx); + + /* wipe the request out so that we can drop the connection early + * if the client closes first. + */ + channel_dont_connect(req); + + req->analysers &= AN_REQ_FLT_END; /* remove switching rules etc... */ + req->analysers |= AN_REQ_HTTP_TARPIT; + req->analyse_exp = tick_add_ifset(now_ms, s->be->timeout.tarpit); + if (!req->analyse_exp) + req->analyse_exp = tick_add(now_ms, 0); + stream_inc_http_err_ctr(s); + _HA_ATOMIC_INC(&sess->fe->fe_counters.denied_req); + if (s->flags & SF_BE_ASSIGNED) + _HA_ATOMIC_INC(&s->be->be_counters.denied_req); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->denied_req); + goto done_without_exp; + + deny: /* this request was blocked (denied) */ + + /* Allow cookie logging + */ + if (s->be->cookie_name || sess->fe->capture_name) + http_manage_client_side_cookies(s, req); + + s->logs.tv_request = now; + stream_inc_http_err_ctr(s); + _HA_ATOMIC_INC(&sess->fe->fe_counters.denied_req); + if (s->flags & SF_BE_ASSIGNED) + _HA_ATOMIC_INC(&s->be->be_counters.denied_req); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->denied_req); + goto return_prx_err; + + return_fail_rewrite: + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + _HA_ATOMIC_INC(&sess->fe->fe_counters.failed_rewrites); + if (s->flags & SF_BE_ASSIGNED) + _HA_ATOMIC_INC(&s->be->be_counters.failed_rewrites); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->failed_rewrites); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_rewrites); + /* fall through */ + + return_int_err: + txn->status = 500; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + _HA_ATOMIC_INC(&sess->fe->fe_counters.internal_errors); + if (s->flags & SF_BE_ASSIGNED) + _HA_ATOMIC_INC(&s->be->be_counters.internal_errors); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->internal_errors); + goto return_prx_err; + + return_bad_req: + txn->status = 400; + _HA_ATOMIC_INC(&sess->fe->fe_counters.failed_req); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->failed_req); + /* fall through */ + + return_prx_err: + http_reply_and_close(s, txn->status, http_error_message(s)); + /* fall through */ + + return_prx_cond: + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_R; + + req->analysers &= AN_REQ_FLT_END; + req->analyse_exp = TICK_ETERNITY; + s->current_rule = s->current_rule_list = NULL; + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + + return_prx_yield: + channel_dont_connect(req); + DBG_TRACE_DEVEL("waiting for more data", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; +} + +/* This function performs all the processing enabled for the current request. + * It returns 1 if the processing can continue on next analysers, or zero if it + * needs more data, encounters an error, or wants to immediately abort the + * request. It relies on buffers flags, and updates s->req.analysers. + */ +int http_process_request(struct stream *s, struct channel *req, int an_bit) +{ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct htx *htx; + struct connection *cli_conn = objt_conn(strm_sess(s)->origin); + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + + /* + * Right now, we know that we have processed the entire headers + * and that unwanted requests have been filtered out. We can do + * whatever we want with the remaining request. Also, now we + * may have separate values for ->fe, ->be. + */ + htx = htxbuf(&req->buf); + + /* + * 7: Now we can work with the cookies. + * Note that doing so might move headers in the request, but + * the fields will stay coherent and the URI will not move. + * This should only be performed in the backend. + */ + if (s->be->cookie_name || sess->fe->capture_name) + http_manage_client_side_cookies(s, req); + + /* 8: Generate unique ID if a "unique-id-format" is defined. + * + * A unique ID is generated even when it is not sent to ensure that the ID can make use of + * fetches only available in the HTTP request processing stage. + */ + if (!LIST_ISEMPTY(&sess->fe->format_unique_id)) { + struct ist unique_id = stream_generate_unique_id(s, &sess->fe->format_unique_id); + + if (!isttest(unique_id)) { + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_RESOURCE; + goto return_int_err; + } + + /* send unique ID if a "unique-id-header" is defined */ + if (isttest(sess->fe->header_unique_id) && + unlikely(!http_add_header(htx, sess->fe->header_unique_id, unique_id))) + goto return_fail_rewrite; + } + + /* + * 9: add X-Forwarded-For if either the frontend or the backend + * asks for it. + */ + if ((sess->fe->options | s->be->options) & PR_O_FWDFOR) { + const struct sockaddr_storage *src = sc_src(s->scf); + struct http_hdr_ctx ctx = { .blk = NULL }; + struct ist hdr = isttest(s->be->fwdfor_hdr_name) ? s->be->fwdfor_hdr_name : sess->fe->fwdfor_hdr_name; + + if (!((sess->fe->options | s->be->options) & PR_O_FF_ALWAYS) && + http_find_header(htx, hdr, &ctx, 0)) { + /* The header is set to be added only if none is present + * and we found it, so don't do anything. + */ + } + else if (src && src->ss_family == AF_INET) { + /* Add an X-Forwarded-For header unless the source IP is + * in the 'except' network range. + */ + if (ipcmp2net(src, &sess->fe->except_xff_net) && + ipcmp2net(src, &s->be->except_xff_net)) { + unsigned char *pn = (unsigned char *)&((struct sockaddr_in *)src)->sin_addr; + + /* Note: we rely on the backend to get the header name to be used for + * x-forwarded-for, because the header is really meant for the backends. + * However, if the backend did not specify any option, we have to rely + * on the frontend's header name. + */ + chunk_printf(&trash, "%d.%d.%d.%d", pn[0], pn[1], pn[2], pn[3]); + if (unlikely(!http_add_header(htx, hdr, ist2(trash.area, trash.data)))) + goto return_fail_rewrite; + } + } + else if (src && src->ss_family == AF_INET6) { + /* Add an X-Forwarded-For header unless the source IP is + * in the 'except' network range. + */ + if (ipcmp2net(src, &sess->fe->except_xff_net) && + ipcmp2net(src, &s->be->except_xff_net)) { + char pn[INET6_ADDRSTRLEN]; + + inet_ntop(AF_INET6, + (const void *)&((struct sockaddr_in6 *)(src))->sin6_addr, + pn, sizeof(pn)); + + /* Note: we rely on the backend to get the header name to be used for + * x-forwarded-for, because the header is really meant for the backends. + * However, if the backend did not specify any option, we have to rely + * on the frontend's header name. + */ + chunk_printf(&trash, "%s", pn); + if (unlikely(!http_add_header(htx, hdr, ist2(trash.area, trash.data)))) + goto return_fail_rewrite; + } + } + } + + /* + * 10: add X-Original-To if either the frontend or the backend + * asks for it. + */ + if ((sess->fe->options | s->be->options) & PR_O_ORGTO) { + const struct sockaddr_storage *dst = sc_dst(s->scf); + struct ist hdr = isttest(s->be->orgto_hdr_name) ? s->be->orgto_hdr_name : sess->fe->orgto_hdr_name; + + if (dst && dst->ss_family == AF_INET) { + /* Add an X-Original-To header unless the destination IP is + * in the 'except' network range. + */ + if (ipcmp2net(dst, &sess->fe->except_xot_net) && + ipcmp2net(dst, &s->be->except_xot_net)) { + unsigned char *pn = (unsigned char *)&((struct sockaddr_in *)dst)->sin_addr; + + /* Note: we rely on the backend to get the header name to be used for + * x-original-to, because the header is really meant for the backends. + * However, if the backend did not specify any option, we have to rely + * on the frontend's header name. + */ + chunk_printf(&trash, "%d.%d.%d.%d", pn[0], pn[1], pn[2], pn[3]); + if (unlikely(!http_add_header(htx, hdr, ist2(trash.area, trash.data)))) + goto return_fail_rewrite; + } + } + else if (dst && dst->ss_family == AF_INET6) { + /* Add an X-Original-To header unless the source IP is + * in the 'except' network range. + */ + if (ipcmp2net(dst, &sess->fe->except_xot_net) && + ipcmp2net(dst, &s->be->except_xot_net)) { + char pn[INET6_ADDRSTRLEN]; + + inet_ntop(AF_INET6, + (const void *)&((struct sockaddr_in6 *)dst)->sin6_addr, + pn, sizeof(pn)); + + /* Note: we rely on the backend to get the header name to be used for + * x-forwarded-for, because the header is really meant for the backends. + * However, if the backend did not specify any option, we have to rely + * on the frontend's header name. + */ + chunk_printf(&trash, "%s", pn); + if (unlikely(!http_add_header(htx, hdr, ist2(trash.area, trash.data)))) + goto return_fail_rewrite; + } + } + } + + /* Filter the request headers if there are filters attached to the + * stream. + */ + if (HAS_FILTERS(s)) + req->analysers |= AN_REQ_FLT_HTTP_HDRS; + + /* If we have no server assigned yet and we're balancing on url_param + * with a POST request, we may be interested in checking the body for + * that parameter. This will be done in another analyser. + */ + if (!(s->flags & (SF_ASSIGNED|SF_DIRECT)) && + s->txn->meth == HTTP_METH_POST && + (s->be->lbprm.algo & BE_LB_ALGO) == BE_LB_ALGO_PH) { + channel_dont_connect(req); + req->analysers |= AN_REQ_HTTP_BODY; + } + + req->analysers &= ~AN_REQ_FLT_XFER_DATA; + req->analysers |= AN_REQ_HTTP_XFER_BODY; + + /* We expect some data from the client. Unless we know for sure + * we already have a full request, we have to re-enable quick-ack + * in case we previously disabled it, otherwise we might cause + * the client to delay further data. + */ + if ((sess->listener && (sess->listener->options & LI_O_NOQUICKACK)) && !(htx->flags & HTX_FL_EOM)) + conn_set_quickack(cli_conn, 1); + + /************************************************************* + * OK, that's finished for the headers. We have done what we * + * could. Let's switch to the DATA state. * + ************************************************************/ + req->analyse_exp = TICK_ETERNITY; + req->analysers &= ~an_bit; + + s->logs.tv_request = now; + /* OK let's go on with the BODY now */ + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 1; + + return_fail_rewrite: + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + _HA_ATOMIC_INC(&sess->fe->fe_counters.failed_rewrites); + if (s->flags & SF_BE_ASSIGNED) + _HA_ATOMIC_INC(&s->be->be_counters.failed_rewrites); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->failed_rewrites); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_rewrites); + /* fall through */ + + return_int_err: + txn->status = 500; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + _HA_ATOMIC_INC(&sess->fe->fe_counters.internal_errors); + if (s->flags & SF_BE_ASSIGNED) + _HA_ATOMIC_INC(&s->be->be_counters.internal_errors); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->internal_errors); + + http_reply_and_close(s, txn->status, http_error_message(s)); + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_R; + + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; +} + +/* This function is an analyser which processes the HTTP tarpit. It always + * returns zero, at the beginning because it prevents any other processing + * from occurring, and at the end because it terminates the request. + */ +int http_process_tarpit(struct stream *s, struct channel *req, int an_bit) +{ + struct http_txn *txn = s->txn; + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn, &txn->req); + /* This connection is being tarpitted. The CLIENT side has + * already set the connect expiration date to the right + * timeout. We just have to check that the client is still + * there and that the timeout has not expired. + */ + channel_dont_connect(req); + if ((req->flags & (CF_SHUTR|CF_READ_ERROR)) == 0 && + !tick_is_expired(req->analyse_exp, now_ms)) { + /* Be sure to drain all data from the request channel */ + channel_htx_erase(req, htxbuf(&req->buf)); + DBG_TRACE_DEVEL("waiting for tarpit timeout expiry", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + } + + + /* We will set the queue timer to the time spent, just for + * logging purposes. We fake a 500 server error, so that the + * attacker will not suspect his connection has been tarpitted. + * It will not cause trouble to the logs because we can exclude + * the tarpitted connections by filtering on the 'PT' status flags. + */ + s->logs.t_queue = tv_ms_elapsed(&s->logs.tv_accept, &now); + + http_reply_and_close(s, txn->status, (!(req->flags & CF_READ_ERROR) ? http_error_message(s) : NULL)); + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_T; + + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; +} + +/* This function is an analyser which waits for the HTTP request body. It waits + * for either the buffer to be full, or the full advertised contents to have + * reached the buffer. It must only be called after the standard HTTP request + * processing has occurred, because it expects the request to be parsed and will + * look for the Expect header. It may send a 100-Continue interim response. It + * takes in input any state starting from HTTP_MSG_BODY and leaves with one of + * HTTP_MSG_CHK_SIZE, HTTP_MSG_DATA or HTTP_MSG_TRAILERS. It returns zero if it + * needs to read more data, or 1 once it has completed its analysis. + */ +int http_wait_for_request_body(struct stream *s, struct channel *req, int an_bit) +{ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn, &s->txn->req); + + + switch (http_wait_for_msg_body(s, req, s->be->timeout.httpreq, 0)) { + case HTTP_RULE_RES_CONT: + goto http_end; + case HTTP_RULE_RES_YIELD: + goto missing_data_or_waiting; + case HTTP_RULE_RES_BADREQ: + goto return_bad_req; + case HTTP_RULE_RES_ERROR: + goto return_int_err; + case HTTP_RULE_RES_ABRT: + goto return_prx_cond; + default: + goto return_int_err; + } + + http_end: + /* The situation will not evolve, so let's give up on the analysis. */ + s->logs.tv_request = now; /* update the request timer to reflect full request */ + req->analysers &= ~an_bit; + req->analyse_exp = TICK_ETERNITY; + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 1; + + missing_data_or_waiting: + channel_dont_connect(req); + DBG_TRACE_DEVEL("waiting for more data", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + + return_int_err: + txn->status = 500; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + _HA_ATOMIC_INC(&sess->fe->fe_counters.internal_errors); + if (s->flags & SF_BE_ASSIGNED) + _HA_ATOMIC_INC(&s->be->be_counters.internal_errors); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->internal_errors); + goto return_prx_err; + + return_bad_req: /* let's centralize all bad requests */ + txn->status = 400; + _HA_ATOMIC_INC(&sess->fe->fe_counters.failed_req); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->failed_req); + /* fall through */ + + return_prx_err: + http_reply_and_close(s, txn->status, http_error_message(s)); + /* fall through */ + + return_prx_cond: + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_R; + + req->analysers &= AN_REQ_FLT_END; + req->analyse_exp = TICK_ETERNITY; + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; +} + +/* This function is an analyser which forwards request body (including chunk + * sizes if any). It is called as soon as we must forward, even if we forward + * zero byte. The only situation where it must not be called is when we're in + * tunnel mode and we want to forward till the close. It's used both to forward + * remaining data and to resync after end of body. It expects the msg_state to + * be between MSG_BODY and MSG_DONE (inclusive). It returns zero if it needs to + * read more data, or 1 once we can go on with next request or end the stream. + * When in MSG_DATA or MSG_TRAILERS, it will automatically forward chunk_len + * bytes of pending data + the headers if not already done. + */ +int http_request_forward_body(struct stream *s, struct channel *req, int an_bit) +{ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct http_msg *msg = &txn->req; + struct htx *htx; + short status = 0; + int ret; + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn, msg); + + htx = htxbuf(&req->buf); + + if (htx->flags & HTX_FL_PARSING_ERROR) + goto return_bad_req; + if (htx->flags & HTX_FL_PROCESSING_ERROR) + goto return_int_err; + + if ((req->flags & (CF_READ_ERROR|CF_READ_TIMEOUT|CF_WRITE_ERROR|CF_WRITE_TIMEOUT)) || + ((req->flags & CF_SHUTW) && (req->to_forward || co_data(req)))) { + /* Output closed while we were sending data. We must abort and + * wake the other side up. + * + * If we have finished to send the request and the response is + * still in progress, don't catch write error on the request + * side if it is in fact a read error on the server side. + */ + if (msg->msg_state == HTTP_MSG_DONE && (s->res.flags & CF_READ_ERROR) && s->res.analysers) + return 0; + + /* Don't abort yet if we had L7 retries activated and it + * was a write error, we may recover. + */ + if (!(req->flags & (CF_READ_ERROR | CF_READ_TIMEOUT)) && + (txn->flags & TX_L7_RETRY)) { + DBG_TRACE_DEVEL("leaving on L7 retry", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + } + msg->msg_state = HTTP_MSG_ERROR; + http_end_request(s); + http_end_response(s); + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 1; + } + + /* Note that we don't have to send 100-continue back because we don't + * need the data to complete our job, and it's up to the server to + * decide whether to return 100, 417 or anything else in return of + * an "Expect: 100-continue" header. + */ + if (msg->msg_state == HTTP_MSG_BODY) + msg->msg_state = HTTP_MSG_DATA; + + /* in most states, we should abort in case of early close */ + channel_auto_close(req); + + if (req->to_forward) { + if (req->to_forward == CHN_INFINITE_FORWARD) { + if (req->flags & CF_EOI) + msg->msg_state = HTTP_MSG_ENDING; + } + else { + /* We can't process the buffer's contents yet */ + req->flags |= CF_WAKE_WRITE; + goto missing_data_or_waiting; + } + } + + if (msg->msg_state >= HTTP_MSG_ENDING) + goto ending; + + if (txn->meth == HTTP_METH_CONNECT) { + msg->msg_state = HTTP_MSG_ENDING; + goto ending; + } + + /* Forward input data. We get it by removing all outgoing data not + * forwarded yet from HTX data size. If there are some data filters, we + * let them decide the amount of data to forward. + */ + if (HAS_REQ_DATA_FILTERS(s)) { + ret = flt_http_payload(s, msg, htx->data); + if (ret < 0) + goto return_bad_req; + c_adv(req, ret); + } + else { + c_adv(req, htx->data - co_data(req)); + if (msg->flags & HTTP_MSGF_XFER_LEN) + channel_htx_forward_forever(req, htx); + } + + if (htx->data != co_data(req)) + goto missing_data_or_waiting; + + /* Check if the end-of-message is reached and if so, switch the message + * in HTTP_MSG_ENDING state. Then if all data was marked to be + * forwarded, set the state to HTTP_MSG_DONE. + */ + if (!(htx->flags & HTX_FL_EOM)) + goto missing_data_or_waiting; + + msg->msg_state = HTTP_MSG_ENDING; + + ending: + req->flags &= ~CF_EXPECT_MORE; /* no more data are expected */ + + /* other states, ENDING...TUNNEL */ + if (msg->msg_state >= HTTP_MSG_DONE) + goto done; + + if (HAS_REQ_DATA_FILTERS(s)) { + ret = flt_http_end(s, msg); + if (ret <= 0) { + if (!ret) + goto missing_data_or_waiting; + goto return_bad_req; + } + } + + if (txn->meth == HTTP_METH_CONNECT) + msg->msg_state = HTTP_MSG_TUNNEL; + else { + msg->msg_state = HTTP_MSG_DONE; + req->to_forward = 0; + } + + done: + /* we don't want to forward closes on DONE except in tunnel mode. */ + if (!(txn->flags & TX_CON_WANT_TUN)) + channel_dont_close(req); + + http_end_request(s); + if (!(req->analysers & an_bit)) { + http_end_response(s); + if (unlikely(msg->msg_state == HTTP_MSG_ERROR)) { + if (req->flags & CF_SHUTW) { + /* request errors are most likely due to the + * server aborting the transfer. */ + goto return_srv_abort; + } + goto return_bad_req; + } + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 1; + } + + /* If "option abortonclose" is set on the backend, we want to monitor + * the client's connection and forward any shutdown notification to the + * server, which will decide whether to close or to go on processing the + * request. We only do that in tunnel mode, and not in other modes since + * it can be abused to exhaust source ports. */ + if (s->be->options & PR_O_ABRT_CLOSE) { + channel_auto_read(req); + if ((req->flags & (CF_SHUTR|CF_READ_NULL)) && !(txn->flags & TX_CON_WANT_TUN)) + s->scb->flags |= SC_FL_NOLINGER; + channel_auto_close(req); + } + else if (s->txn->meth == HTTP_METH_POST) { + /* POST requests may require to read extra CRLF sent by broken + * browsers and which could cause an RST to be sent upon close + * on some systems (eg: Linux). */ + channel_auto_read(req); + } + DBG_TRACE_DEVEL("waiting for the end of the HTTP txn", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + + missing_data_or_waiting: + /* stop waiting for data if the input is closed before the end */ + if (msg->msg_state < HTTP_MSG_ENDING && req->flags & CF_SHUTR) + goto return_cli_abort; + + waiting: + /* waiting for the last bits to leave the buffer */ + if (req->flags & CF_SHUTW) + goto return_srv_abort; + + /* When TE: chunked is used, we need to get there again to parse remaining + * chunks even if the client has closed, so we don't want to set CF_DONTCLOSE. + * And when content-length is used, we never want to let the possible + * shutdown be forwarded to the other side, as the state machine will + * take care of it once the client responds. It's also important to + * prevent TIME_WAITs from accumulating on the backend side, and for + * HTTP/2 where the last frame comes with a shutdown. + */ + if (msg->flags & HTTP_MSGF_XFER_LEN) + channel_dont_close(req); + + /* We know that more data are expected, but we couldn't send more that + * what we did. So we always set the CF_EXPECT_MORE flag so that the + * system knows it must not set a PUSH on this first part. Interactive + * modes are already handled by the stream sock layer. We must not do + * this in content-length mode because it could present the MSG_MORE + * flag with the last block of forwarded data, which would cause an + * additional delay to be observed by the receiver. + */ + if (HAS_REQ_DATA_FILTERS(s)) + req->flags |= CF_EXPECT_MORE; + + DBG_TRACE_DEVEL("waiting for more data to forward", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + + return_cli_abort: + _HA_ATOMIC_INC(&sess->fe->fe_counters.cli_aborts); + _HA_ATOMIC_INC(&s->be->be_counters.cli_aborts); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->cli_aborts); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.cli_aborts); + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_CLICL; + status = 400; + goto return_prx_cond; + + return_srv_abort: + _HA_ATOMIC_INC(&sess->fe->fe_counters.srv_aborts); + _HA_ATOMIC_INC(&s->be->be_counters.srv_aborts); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->srv_aborts); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.srv_aborts); + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_SRVCL; + status = 502; + goto return_prx_cond; + + return_int_err: + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + _HA_ATOMIC_INC(&sess->fe->fe_counters.internal_errors); + _HA_ATOMIC_INC(&s->be->be_counters.internal_errors); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->internal_errors); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.internal_errors); + status = 500; + goto return_prx_cond; + + return_bad_req: + _HA_ATOMIC_INC(&sess->fe->fe_counters.failed_req); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->failed_req); + status = 400; + /* fall through */ + + return_prx_cond: + if (txn->status > 0) { + /* Note: we don't send any error if some data were already sent */ + http_reply_and_close(s, txn->status, NULL); + } else { + txn->status = status; + http_reply_and_close(s, txn->status, http_error_message(s)); + } + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= ((txn->rsp.msg_state < HTTP_MSG_ERROR) ? SF_FINST_H : SF_FINST_D); + DBG_TRACE_DEVEL("leaving on error ", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; +} + +/* Reset the stream and the backend stream connector to a situation suitable for attemption connection */ +/* Returns 0 if we can attempt to retry, -1 otherwise */ +static __inline int do_l7_retry(struct stream *s, struct stconn *sc) +{ + struct channel *req, *res; + int co_data; + + if (s->conn_retries >= s->be->conn_retries) + return -1; + s->conn_retries++; + if (objt_server(s->target)) { + if (s->flags & SF_CURR_SESS) { + s->flags &= ~SF_CURR_SESS; + _HA_ATOMIC_DEC(&__objt_server(s->target)->cur_sess); + } + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.retries); + } + _HA_ATOMIC_INC(&s->be->be_counters.retries); + + req = &s->req; + res = &s->res; + /* Remove any write error from the request, and read error from the response */ + req->flags &= ~(CF_WRITE_ERROR | CF_WRITE_TIMEOUT | CF_SHUTW | CF_SHUTW_NOW); + res->flags &= ~(CF_READ_ERROR | CF_READ_TIMEOUT | CF_SHUTR | CF_EOI | CF_READ_NULL | CF_SHUTR_NOW); + res->analysers &= AN_RES_FLT_END; + s->conn_err_type = STRM_ET_NONE; + s->flags &= ~(SF_CONN_EXP | SF_ERR_MASK | SF_FINST_MASK); + s->conn_exp = TICK_ETERNITY; + stream_choose_redispatch(s); + res->rex = TICK_ETERNITY; + res->to_forward = 0; + res->analyse_exp = TICK_ETERNITY; + res->total = 0; + + if (sc_reset_endp(s->scb) < 0) { + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + return -1; + } + + b_free(&req->buf); + /* Swap the L7 buffer with the channel buffer */ + /* We know we stored the co_data as b_data, so get it there */ + co_data = b_data(&s->txn->l7_buffer); + b_set_data(&s->txn->l7_buffer, b_size(&s->txn->l7_buffer)); + b_xfer(&req->buf, &s->txn->l7_buffer, b_data(&s->txn->l7_buffer)); + co_set_data(req, co_data); + + DBG_TRACE_DEVEL("perform a L7 retry", STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, s->txn); + + b_reset(&res->buf); + co_set_data(res, 0); + return 0; +} + +/* This stream analyser waits for a complete HTTP response. It returns 1 if the + * processing can continue on next analysers, or zero if it either needs more + * data or wants to immediately abort the response (eg: timeout, error, ...). It + * is tied to AN_RES_WAIT_HTTP and may may remove itself from s->res.analysers + * when it has nothing left to do, and may remove any analyser when it wants to + * abort. + */ +int http_wait_for_response(struct stream *s, struct channel *rep, int an_bit) +{ + /* + * We will analyze a complete HTTP response to check the its syntax. + * + * Once the start line and all headers are received, we may perform a + * capture of the error (if any), and we will set a few fields. We also + * logging and finally headers capture. + */ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct http_msg *msg = &txn->rsp; + struct htx *htx; + struct connection *srv_conn; + struct htx_sl *sl; + int n; + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn, msg); + + htx = htxbuf(&rep->buf); + + /* Parsing errors are caught here */ + if (htx->flags & HTX_FL_PARSING_ERROR) + goto return_bad_res; + if (htx->flags & HTX_FL_PROCESSING_ERROR) + goto return_int_err; + + /* + * Now we quickly check if we have found a full valid response. + * If not so, we check the FD and buffer states before leaving. + * A full response is indicated by the fact that we have seen + * the double LF/CRLF, so the state is >= HTTP_MSG_BODY. Invalid + * responses are checked first. + * + * Depending on whether the client is still there or not, we + * may send an error response back or not. Note that normally + * we should only check for HTTP status there, and check I/O + * errors somewhere else. + */ + next_one: + if (unlikely(htx_is_empty(htx) || htx->first == -1)) { + /* 1: have we encountered a read error ? */ + if (rep->flags & CF_READ_ERROR) { + struct connection *conn = sc_conn(s->scb); + + + if ((txn->flags & TX_L7_RETRY) && + (s->be->retry_type & PR_RE_DISCONNECTED) && + (!conn || conn->err_code != CO_ER_SSL_EARLY_FAILED)) { + if (co_data(rep) || do_l7_retry(s, s->scb) == 0) + return 0; + } + + /* Perform a L7 retry on empty response or because server refuses the early data. */ + if ((txn->flags & TX_L7_RETRY) && + (s->be->retry_type & PR_RE_EARLY_ERROR) && + conn && conn->err_code == CO_ER_SSL_EARLY_FAILED && + do_l7_retry(s, s->scb) == 0) { + DBG_TRACE_DEVEL("leaving on L7 retry", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + } + + if (txn->flags & TX_NOT_FIRST) + goto abort_keep_alive; + + _HA_ATOMIC_INC(&s->be->be_counters.failed_resp); + if (objt_server(s->target)) { + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_resp); + health_adjust(__objt_server(s->target), HANA_STATUS_HTTP_READ_ERROR); + } + + /* if the server refused the early data, just send a 425 */ + if (conn && conn->err_code == CO_ER_SSL_EARLY_FAILED) + txn->status = 425; + else { + txn->status = 502; + stream_inc_http_fail_ctr(s); + } + + s->scb->flags |= SC_FL_NOLINGER; + http_reply_and_close(s, txn->status, http_error_message(s)); + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_SRVCL; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_H; + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + } + + /* 2: read timeout : return a 504 to the client. */ + else if (rep->flags & CF_READ_TIMEOUT) { + if ((txn->flags & TX_L7_RETRY) && + (s->be->retry_type & PR_RE_TIMEOUT)) { + if (co_data(rep) || do_l7_retry(s, s->scb) == 0) { + DBG_TRACE_DEVEL("leaving on L7 retry", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + } + } + _HA_ATOMIC_INC(&s->be->be_counters.failed_resp); + if (objt_server(s->target)) { + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_resp); + health_adjust(__objt_server(s->target), HANA_STATUS_HTTP_READ_TIMEOUT); + } + + txn->status = 504; + stream_inc_http_fail_ctr(s); + s->scb->flags |= SC_FL_NOLINGER; + http_reply_and_close(s, txn->status, http_error_message(s)); + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_SRVTO; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_H; + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + } + + /* 3: client abort with an abortonclose */ + else if ((rep->flags & CF_SHUTR) && ((s->req.flags & (CF_SHUTR|CF_SHUTW)) == (CF_SHUTR|CF_SHUTW))) { + _HA_ATOMIC_INC(&sess->fe->fe_counters.cli_aborts); + _HA_ATOMIC_INC(&s->be->be_counters.cli_aborts); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->cli_aborts); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.cli_aborts); + + txn->status = 400; + http_reply_and_close(s, txn->status, http_error_message(s)); + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_CLICL; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_H; + + /* process_stream() will take care of the error */ + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + } + + /* 4: close from server, capture the response if the server has started to respond */ + else if (rep->flags & CF_SHUTR) { + if ((txn->flags & TX_L7_RETRY) && + (s->be->retry_type & PR_RE_DISCONNECTED)) { + if (co_data(rep) || do_l7_retry(s, s->scb) == 0) { + DBG_TRACE_DEVEL("leaving on L7 retry", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + } + } + + if (txn->flags & TX_NOT_FIRST) + goto abort_keep_alive; + + _HA_ATOMIC_INC(&s->be->be_counters.failed_resp); + if (objt_server(s->target)) { + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_resp); + health_adjust(__objt_server(s->target), HANA_STATUS_HTTP_BROKEN_PIPE); + } + + txn->status = 502; + stream_inc_http_fail_ctr(s); + s->scb->flags |= SC_FL_NOLINGER; + http_reply_and_close(s, txn->status, http_error_message(s)); + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_SRVCL; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_H; + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + } + + /* 5: write error to client (we don't send any message then) */ + else if (rep->flags & CF_WRITE_ERROR) { + if (txn->flags & TX_NOT_FIRST) + goto abort_keep_alive; + + _HA_ATOMIC_INC(&s->be->be_counters.failed_resp); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_resp); + rep->analysers &= AN_RES_FLT_END; + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_CLICL; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_H; + + /* process_stream() will take care of the error */ + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + } + + channel_dont_close(rep); + rep->flags |= CF_READ_DONTWAIT; /* try to get back here ASAP */ + DBG_TRACE_DEVEL("waiting for more data", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + } + + /* More interesting part now : we know that we have a complete + * response which at least looks like HTTP. We have an indicator + * of each header's length, so we can parse them quickly. + */ + BUG_ON(htx_get_first_type(htx) != HTX_BLK_RES_SL); + sl = http_get_stline(htx); + + /* Perform a L7 retry because of the status code */ + if ((txn->flags & TX_L7_RETRY) && + l7_status_match(s->be, sl->info.res.status) && + do_l7_retry(s, s->scb) == 0) { + DBG_TRACE_DEVEL("leaving on L7 retry", STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + } + + /* Now, L7 buffer is useless, it can be released */ + b_free(&txn->l7_buffer); + + msg->msg_state = HTTP_MSG_BODY; + + + /* 0: we might have to print this header in debug mode */ + if (unlikely((global.mode & MODE_DEBUG) && + (!(global.mode & MODE_QUIET) || (global.mode & MODE_VERBOSE)))) { + int32_t pos; + + http_debug_stline("srvrep", s, sl); + + for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) { + struct htx_blk *blk = htx_get_blk(htx, pos); + enum htx_blk_type type = htx_get_blk_type(blk); + + if (type == HTX_BLK_EOH) + break; + if (type != HTX_BLK_HDR) + continue; + + http_debug_hdr("srvhdr", s, + htx_get_blk_name(htx, blk), + htx_get_blk_value(htx, blk)); + } + } + + /* 1: get the status code and the version. Also set HTTP flags */ + txn->status = sl->info.res.status; + if (sl->flags & HTX_SL_F_VER_11) + msg->flags |= HTTP_MSGF_VER_11; + if (sl->flags & HTX_SL_F_XFER_LEN) { + msg->flags |= HTTP_MSGF_XFER_LEN; + if (sl->flags & HTX_SL_F_CLEN) + msg->flags |= HTTP_MSGF_CNT_LEN; + else if (sl->flags & HTX_SL_F_CHNK) + msg->flags |= HTTP_MSGF_TE_CHNK; + } + if (sl->flags & HTX_SL_F_BODYLESS) + msg->flags |= HTTP_MSGF_BODYLESS; + if (sl->flags & HTX_SL_F_CONN_UPG) + msg->flags |= HTTP_MSGF_CONN_UPG; + + n = txn->status / 100; + if (n < 1 || n > 5) + n = 0; + + /* when the client triggers a 4xx from the server, it's most often due + * to a missing object or permission. These events should be tracked + * because if they happen often, it may indicate a brute force or a + * vulnerability scan. + */ + if (n == 4) + stream_inc_http_err_ctr(s); + + if (n == 5 && txn->status != 501 && txn->status != 505) + stream_inc_http_fail_ctr(s); + + if (objt_server(s->target)) { + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.p.http.rsp[n]); + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.p.http.cum_req); + } + + /* Adjust server's health based on status code. Note: status codes 501 + * and 505 are triggered on demand by client request, so we must not + * count them as server failures. + */ + if (objt_server(s->target)) { + if (txn->status >= 100 && (txn->status < 500 || txn->status == 501 || txn->status == 505)) + health_adjust(__objt_server(s->target), HANA_STATUS_HTTP_OK); + else + health_adjust(__objt_server(s->target), HANA_STATUS_HTTP_STS); + } + + /* + * We may be facing a 100-continue response, or any other informational + * 1xx response which is non-final, in which case this is not the right + * response, and we're waiting for the next one. Let's allow this response + * to go to the client and wait for the next one. There's an exception for + * 101 which is used later in the code to switch protocols. + */ + if (txn->status < 200 && + (txn->status == 100 || txn->status >= 102)) { + FLT_STRM_CB(s, flt_http_reset(s, msg)); + htx->first = channel_htx_fwd_headers(rep, htx); + msg->msg_state = HTTP_MSG_RPBEFORE; + msg->flags = 0; + txn->status = 0; + s->logs.t_data = -1; /* was not a response yet */ + rep->flags |= CF_SEND_DONTWAIT; /* Send ASAP informational messages */ + goto next_one; + } + + /* A 101-switching-protocols must contains a Connection header with the + * "upgrade" option and the request too. It means both are agree to + * upgrade. It is not so strict because there is no test on the Upgrade + * header content. But it is probably stronger enough for now. + */ + if (txn->status == 101 && + (!(txn->req.flags & HTTP_MSGF_CONN_UPG) || !(txn->rsp.flags & HTTP_MSGF_CONN_UPG))) + goto return_bad_res; + + /* + * 2: check for cacheability. + */ + + switch (txn->status) { + case 200: + case 203: + case 204: + case 206: + case 300: + case 301: + case 404: + case 405: + case 410: + case 414: + case 501: + break; + default: + /* RFC7231#6.1: + * Responses with status codes that are defined as + * cacheable by default (e.g., 200, 203, 204, 206, + * 300, 301, 404, 405, 410, 414, and 501 in this + * specification) can be reused by a cache with + * heuristic expiration unless otherwise indicated + * by the method definition or explicit cache + * controls [RFC7234]; all other status codes are + * not cacheable by default. + */ + txn->flags &= ~(TX_CACHEABLE | TX_CACHE_COOK); + break; + } + + /* + * 3: we may need to capture headers + */ + s->logs.logwait &= ~LW_RESP; + if (unlikely((s->logs.logwait & LW_RSPHDR) && s->res_cap)) + http_capture_headers(htx, s->res_cap, sess->fe->rsp_cap); + + /* Skip parsing if no content length is possible. */ + if (unlikely((txn->meth == HTTP_METH_CONNECT && txn->status >= 200 && txn->status < 300) || + txn->status == 101)) { + /* Either we've established an explicit tunnel, or we're + * switching the protocol. In both cases, we're very unlikely + * to understand the next protocols. We have to switch to tunnel + * mode, so that we transfer the request and responses then let + * this protocol pass unmodified. When we later implement specific + * parsers for such protocols, we'll want to check the Upgrade + * header which contains information about that protocol for + * responses with status 101 (eg: see RFC2817 about TLS). + */ + txn->flags |= TX_CON_WANT_TUN; + } + + /* check for NTML authentication headers in 401 (WWW-Authenticate) and + * 407 (Proxy-Authenticate) responses and set the connection to private + */ + srv_conn = sc_conn(s->scb); + if (srv_conn) { + struct ist hdr; + struct http_hdr_ctx ctx; + + if (txn->status == 401) + hdr = ist("WWW-Authenticate"); + else if (txn->status == 407) + hdr = ist("Proxy-Authenticate"); + else + goto end; + + ctx.blk = NULL; + while (http_find_header(htx, hdr, &ctx, 0)) { + /* If www-authenticate contains "Negotiate", "Nego2", or "NTLM", + * possibly followed by blanks and a base64 string, the connection + * is private. Since it's a mess to deal with, we only check for + * values starting with "NTLM" or "Nego". Note that often multiple + * headers are sent by the server there. + */ + if ((ctx.value.len >= 4 && strncasecmp(ctx.value.ptr, "Nego", 4) == 0) || + (ctx.value.len >= 4 && strncasecmp(ctx.value.ptr, "NTLM", 4) == 0)) { + sess->flags |= SESS_FL_PREFER_LAST; + conn_set_owner(srv_conn, sess, NULL); + conn_set_private(srv_conn); + /* If it fail now, the same will be done in mux->detach() callback */ + session_add_conn(srv_conn->owner, srv_conn, srv_conn->target); + break; + } + } + } + + end: + /* we want to have the response time before we start processing it */ + s->logs.t_data = tv_ms_elapsed(&s->logs.tv_accept, &now); + + /* end of job, return OK */ + rep->analysers &= ~an_bit; + rep->analyse_exp = TICK_ETERNITY; + channel_auto_close(rep); + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 1; + + return_int_err: + _HA_ATOMIC_INC(&sess->fe->fe_counters.internal_errors); + _HA_ATOMIC_INC(&s->be->be_counters.internal_errors); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->internal_errors); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.internal_errors); + txn->status = 500; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + goto return_prx_cond; + + return_bad_res: + _HA_ATOMIC_INC(&s->be->be_counters.failed_resp); + if (objt_server(s->target)) { + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_resp); + health_adjust(__objt_server(s->target), HANA_STATUS_HTTP_HDRRSP); + } + if ((s->be->retry_type & PR_RE_JUNK_REQUEST) && + (txn->flags & TX_L7_RETRY) && + do_l7_retry(s, s->scb) == 0) { + DBG_TRACE_DEVEL("leaving on L7 retry", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + } + txn->status = 502; + stream_inc_http_fail_ctr(s); + /* fall through */ + + return_prx_cond: + http_reply_and_close(s, txn->status, http_error_message(s)); + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_H; + + s->scb->flags |= SC_FL_NOLINGER; + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + + abort_keep_alive: + /* A keep-alive request to the server failed on a network error. + * The client is required to retry. We need to close without returning + * any other information so that the client retries. + */ + txn->status = 0; + s->logs.logwait = 0; + s->logs.level = 0; + s->res.flags &= ~CF_EXPECT_MORE; /* speed up sending a previous response */ + http_reply_and_close(s, txn->status, NULL); + DBG_TRACE_DEVEL("leaving by closing K/A connection", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; +} + +/* This function performs all the processing enabled for the current response. + * It normally returns 1 unless it wants to break. It relies on buffers flags, + * and updates s->res.analysers. It might make sense to explode it into several + * other functions. It works like process_request (see indications above). + */ +int http_process_res_common(struct stream *s, struct channel *rep, int an_bit, struct proxy *px) +{ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct http_msg *msg = &txn->rsp; + struct htx *htx; + struct proxy *cur_proxy; + enum rule_result ret = HTTP_RULE_RES_CONT; + + if (unlikely(msg->msg_state < HTTP_MSG_BODY)) /* we need more data */ + return 0; + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn, msg); + + htx = htxbuf(&rep->buf); + + /* The stats applet needs to adjust the Connection header but we don't + * apply any filter there. + */ + if (unlikely(objt_applet(s->target) == &http_stats_applet)) { + rep->analysers &= ~an_bit; + rep->analyse_exp = TICK_ETERNITY; + goto end; + } + + /* + * We will have to evaluate the filters. + * As opposed to version 1.2, now they will be evaluated in the + * filters order and not in the header order. This means that + * each filter has to be validated among all headers. + * + * Filters are tried with ->be first, then with ->fe if it is + * different from ->be. + * + * Maybe we are in resume condiion. In this case I choose the + * "struct proxy" which contains the rule list matching the resume + * pointer. If none of these "struct proxy" match, I initialise + * the process with the first one. + * + * In fact, I check only correspondence between the current list + * pointer and the ->fe rule list. If it doesn't match, I initialize + * the loop with the ->be. + */ + if (s->current_rule_list == &sess->fe->http_res_rules || + (sess->fe->defpx && s->current_rule_list == &sess->fe->defpx->http_res_rules)) + cur_proxy = sess->fe; + else + cur_proxy = s->be; + + while (1) { + /* evaluate http-response rules */ + if (ret == HTTP_RULE_RES_CONT || ret == HTTP_RULE_RES_STOP) { + struct list *def_rules, *rules; + + def_rules = ((cur_proxy->defpx && (cur_proxy == s->be || cur_proxy->defpx != s->be->defpx)) ? &cur_proxy->defpx->http_res_rules : NULL); + rules = &cur_proxy->http_res_rules; + + ret = http_res_get_intercept_rule(cur_proxy, def_rules, rules, s); + + switch (ret) { + case HTTP_RULE_RES_YIELD: /* some data miss, call the function later. */ + goto return_prx_yield; + + case HTTP_RULE_RES_CONT: + case HTTP_RULE_RES_STOP: /* nothing to do */ + break; + + case HTTP_RULE_RES_DENY: /* deny or tarpit */ + goto deny; + + case HTTP_RULE_RES_ABRT: /* abort request, response already sent */ + goto return_prx_cond; + + case HTTP_RULE_RES_DONE: /* OK, but terminate request processing (eg: redirect) */ + goto done; + + case HTTP_RULE_RES_BADREQ: /* failed with a bad request */ + goto return_bad_res; + + case HTTP_RULE_RES_ERROR: /* failed with a bad request */ + goto return_int_err; + } + + } + + /* check whether we're already working on the frontend */ + if (cur_proxy == sess->fe) + break; + cur_proxy = sess->fe; + } + + /* OK that's all we can do for 1xx responses */ + if (unlikely(txn->status < 200 && txn->status != 101)) + goto end; + + /* + * Now check for a server cookie. + */ + if (s->be->cookie_name || sess->fe->capture_name || (s->be->options & PR_O_CHK_CACHE)) + http_manage_server_side_cookies(s, rep); + + /* + * Check for cache-control or pragma headers if required. + */ + if ((s->be->options & PR_O_CHK_CACHE) || (s->be->ck_opts & PR_CK_NOC)) + http_check_response_for_cacheability(s, rep); + + /* + * Add server cookie in the response if needed + */ + if (objt_server(s->target) && (s->be->ck_opts & PR_CK_INS) && + !((txn->flags & TX_SCK_FOUND) && (s->be->ck_opts & PR_CK_PSV)) && + (!(s->flags & SF_DIRECT) || + ((s->be->cookie_maxidle || txn->cookie_last_date) && + (!txn->cookie_last_date || (txn->cookie_last_date - date.tv_sec) < 0)) || + (s->be->cookie_maxlife && !txn->cookie_first_date) || // set the first_date + (!s->be->cookie_maxlife && txn->cookie_first_date)) && // remove the first_date + (!(s->be->ck_opts & PR_CK_POST) || (txn->meth == HTTP_METH_POST)) && + !(s->flags & SF_IGNORE_PRST)) { + /* the server is known, it's not the one the client requested, or the + * cookie's last seen date needs to be refreshed. We have to + * insert a set-cookie here, except if we want to insert only on POST + * requests and this one isn't. Note that servers which don't have cookies + * (eg: some backup servers) will return a full cookie removal request. + */ + if (!__objt_server(s->target)->cookie) { + chunk_printf(&trash, + "%s=; Expires=Thu, 01-Jan-1970 00:00:01 GMT; path=/", + s->be->cookie_name); + } + else { + chunk_printf(&trash, "%s=%s", s->be->cookie_name, __objt_server(s->target)->cookie); + + if (s->be->cookie_maxidle || s->be->cookie_maxlife) { + /* emit last_date, which is mandatory */ + trash.area[trash.data++] = COOKIE_DELIM_DATE; + s30tob64((date.tv_sec+3) >> 2, + trash.area + trash.data); + trash.data += 5; + + if (s->be->cookie_maxlife) { + /* emit first_date, which is either the original one or + * the current date. + */ + trash.area[trash.data++] = COOKIE_DELIM_DATE; + s30tob64(txn->cookie_first_date ? + txn->cookie_first_date >> 2 : + (date.tv_sec+3) >> 2, + trash.area + trash.data); + trash.data += 5; + } + } + chunk_appendf(&trash, "; path=/"); + } + + if (s->be->cookie_domain) + chunk_appendf(&trash, "; domain=%s", s->be->cookie_domain); + + if (s->be->ck_opts & PR_CK_HTTPONLY) + chunk_appendf(&trash, "; HttpOnly"); + + if (s->be->ck_opts & PR_CK_SECURE) + chunk_appendf(&trash, "; Secure"); + + if (s->be->cookie_attrs) + chunk_appendf(&trash, "; %s", s->be->cookie_attrs); + + if (unlikely(!http_add_header(htx, ist("Set-Cookie"), ist2(trash.area, trash.data)))) + goto return_fail_rewrite; + + txn->flags &= ~TX_SCK_MASK; + if (__objt_server(s->target)->cookie && (s->flags & SF_DIRECT)) + /* the server did not change, only the date was updated */ + txn->flags |= TX_SCK_UPDATED; + else + txn->flags |= TX_SCK_INSERTED; + + /* Here, we will tell an eventual cache on the client side that we don't + * want it to cache this reply because HTTP/1.0 caches also cache cookies ! + * Some caches understand the correct form: 'no-cache="set-cookie"', but + * others don't (eg: apache <= 1.3.26). So we use 'private' instead. + */ + if ((s->be->ck_opts & PR_CK_NOC) && (txn->flags & TX_CACHEABLE)) { + + txn->flags &= ~TX_CACHEABLE & ~TX_CACHE_COOK; + + if (unlikely(!http_add_header(htx, ist("Cache-control"), ist("private")))) + goto return_fail_rewrite; + } + } + + /* + * Check if result will be cacheable with a cookie. + * We'll block the response if security checks have caught + * nasty things such as a cacheable cookie. + */ + if (((txn->flags & (TX_CACHEABLE | TX_CACHE_COOK | TX_SCK_PRESENT)) == + (TX_CACHEABLE | TX_CACHE_COOK | TX_SCK_PRESENT)) && + (s->be->options & PR_O_CHK_CACHE)) { + /* we're in presence of a cacheable response containing + * a set-cookie header. We'll block it as requested by + * the 'checkcache' option, and send an alert. + */ + ha_alert("Blocking cacheable cookie in response from instance %s, server %s.\n", + s->be->id, objt_server(s->target) ? __objt_server(s->target)->id : "<dispatch>"); + send_log(s->be, LOG_ALERT, + "Blocking cacheable cookie in response from instance %s, server %s.\n", + s->be->id, objt_server(s->target) ? __objt_server(s->target)->id : "<dispatch>"); + goto deny; + } + + end: + /* + * Evaluate after-response rules before forwarding the response. rules + * from the backend are evaluated first, then one from the frontend if + * it differs. + */ + if (!http_eval_after_res_rules(s)) + goto return_int_err; + + /* Filter the response headers if there are filters attached to the + * stream. + */ + if (HAS_FILTERS(s)) + rep->analysers |= AN_RES_FLT_HTTP_HDRS; + + /* Always enter in the body analyzer */ + rep->analysers &= ~AN_RES_FLT_XFER_DATA; + rep->analysers |= AN_RES_HTTP_XFER_BODY; + + /* if the user wants to log as soon as possible, without counting + * bytes from the server, then this is the right moment. We have + * to temporarily assign bytes_out to log what we currently have. + */ + if (!LIST_ISEMPTY(&sess->fe->logformat) && !(s->logs.logwait & LW_BYTES)) { + s->logs.t_close = s->logs.t_data; /* to get a valid end date */ + s->logs.bytes_out = htx->data; + s->do_log(s); + s->logs.bytes_out = 0; + } + + done: + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + rep->analysers &= ~an_bit; + rep->analyse_exp = TICK_ETERNITY; + s->current_rule = s->current_rule_list = NULL; + return 1; + + deny: + _HA_ATOMIC_INC(&sess->fe->fe_counters.denied_resp); + _HA_ATOMIC_INC(&s->be->be_counters.denied_resp); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->denied_resp); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.denied_resp); + goto return_prx_err; + + return_fail_rewrite: + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + _HA_ATOMIC_INC(&sess->fe->fe_counters.failed_rewrites); + _HA_ATOMIC_INC(&s->be->be_counters.failed_rewrites); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->failed_rewrites); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_rewrites); + /* fall through */ + + return_int_err: + txn->status = 500; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + _HA_ATOMIC_INC(&sess->fe->fe_counters.internal_errors); + _HA_ATOMIC_INC(&s->be->be_counters.internal_errors); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->internal_errors); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.internal_errors); + goto return_prx_err; + + return_bad_res: + txn->status = 502; + stream_inc_http_fail_ctr(s); + _HA_ATOMIC_INC(&s->be->be_counters.failed_resp); + if (objt_server(s->target)) { + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_resp); + health_adjust(__objt_server(s->target), HANA_STATUS_HTTP_RSP); + } + /* fall through */ + + return_prx_err: + http_reply_and_close(s, txn->status, http_error_message(s)); + /* fall through */ + + return_prx_cond: + s->logs.t_data = -1; /* was not a valid response */ + s->scb->flags |= SC_FL_NOLINGER; + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_H; + + rep->analysers &= AN_RES_FLT_END; + s->req.analysers &= AN_REQ_FLT_END; + rep->analyse_exp = TICK_ETERNITY; + s->current_rule = s->current_rule_list = NULL; + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; + + return_prx_yield: + channel_dont_close(rep); + DBG_TRACE_DEVEL("waiting for more data", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; +} + +/* This function is an analyser which forwards response body (including chunk + * sizes if any). It is called as soon as we must forward, even if we forward + * zero byte. The only situation where it must not be called is when we're in + * tunnel mode and we want to forward till the close. It's used both to forward + * remaining data and to resync after end of body. It expects the msg_state to + * be between MSG_BODY and MSG_DONE (inclusive). It returns zero if it needs to + * read more data, or 1 once we can go on with next request or end the stream. + * + * It is capable of compressing response data both in content-length mode and + * in chunked mode. The state machines follows different flows depending on + * whether content-length and chunked modes are used, since there are no + * trailers in content-length : + * + * chk-mode cl-mode + * ,----- BODY -----. + * / \ + * V size > 0 V chk-mode + * .--> SIZE -------------> DATA -------------> CRLF + * | | size == 0 | last byte | + * | v final crlf v inspected | + * | TRAILERS -----------> DONE | + * | | + * `----------------------------------------------' + * + * Compression only happens in the DATA state, and must be flushed in final + * states (TRAILERS/DONE) or when leaving on missing data. Normal forwarding + * is performed at once on final states for all bytes parsed, or when leaving + * on missing data. + */ +int http_response_forward_body(struct stream *s, struct channel *res, int an_bit) +{ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct http_msg *msg = &s->txn->rsp; + struct htx *htx; + int ret; + + DBG_TRACE_ENTER(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn, msg); + + htx = htxbuf(&res->buf); + + if (htx->flags & HTX_FL_PARSING_ERROR) + goto return_bad_res; + if (htx->flags & HTX_FL_PROCESSING_ERROR) + goto return_int_err; + + if ((res->flags & (CF_READ_ERROR|CF_READ_TIMEOUT|CF_WRITE_ERROR|CF_WRITE_TIMEOUT)) || + ((res->flags & CF_SHUTW) && (res->to_forward || co_data(res)))) { + /* Output closed while we were sending data. We must abort and + * wake the other side up. + */ + msg->msg_state = HTTP_MSG_ERROR; + http_end_response(s); + http_end_request(s); + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 1; + } + + if (msg->msg_state == HTTP_MSG_BODY) + msg->msg_state = HTTP_MSG_DATA; + + /* in most states, we should abort in case of early close */ + channel_auto_close(res); + + if (res->to_forward) { + if (res->to_forward == CHN_INFINITE_FORWARD) { + if (res->flags & CF_EOI) + msg->msg_state = HTTP_MSG_ENDING; + } + else { + /* We can't process the buffer's contents yet */ + res->flags |= CF_WAKE_WRITE; + goto missing_data_or_waiting; + } + } + + if (msg->msg_state >= HTTP_MSG_ENDING) + goto ending; + + if ((txn->meth == HTTP_METH_CONNECT && txn->status >= 200 && txn->status < 300) || txn->status == 101 || + (!(msg->flags & HTTP_MSGF_XFER_LEN) && !HAS_RSP_DATA_FILTERS(s))) { + msg->msg_state = HTTP_MSG_ENDING; + goto ending; + } + + /* Forward input data. We get it by removing all outgoing data not + * forwarded yet from HTX data size. If there are some data filters, we + * let them decide the amount of data to forward. + */ + if (HAS_RSP_DATA_FILTERS(s)) { + ret = flt_http_payload(s, msg, htx->data); + if (ret < 0) + goto return_bad_res; + c_adv(res, ret); + } + else { + c_adv(res, htx->data - co_data(res)); + if (msg->flags & HTTP_MSGF_XFER_LEN) + channel_htx_forward_forever(res, htx); + } + + if (htx->data != co_data(res)) + goto missing_data_or_waiting; + + if (!(msg->flags & HTTP_MSGF_XFER_LEN) && res->flags & CF_SHUTR) { + msg->msg_state = HTTP_MSG_ENDING; + goto ending; + } + + /* Check if the end-of-message is reached and if so, switch the message + * in HTTP_MSG_ENDING state. Then if all data was marked to be + * forwarded, set the state to HTTP_MSG_DONE. + */ + if (!(htx->flags & HTX_FL_EOM)) + goto missing_data_or_waiting; + + msg->msg_state = HTTP_MSG_ENDING; + + ending: + res->flags &= ~CF_EXPECT_MORE; /* no more data are expected */ + + /* other states, ENDING...TUNNEL */ + if (msg->msg_state >= HTTP_MSG_DONE) + goto done; + + if (HAS_RSP_DATA_FILTERS(s)) { + ret = flt_http_end(s, msg); + if (ret <= 0) { + if (!ret) + goto missing_data_or_waiting; + goto return_bad_res; + } + } + + if ((txn->meth == HTTP_METH_CONNECT && txn->status >= 200 && txn->status < 300) || txn->status == 101 || + !(msg->flags & HTTP_MSGF_XFER_LEN)) { + msg->msg_state = HTTP_MSG_TUNNEL; + goto ending; + } + else { + msg->msg_state = HTTP_MSG_DONE; + res->to_forward = 0; + } + + done: + + channel_dont_close(res); + + http_end_response(s); + if (!(res->analysers & an_bit)) { + http_end_request(s); + if (unlikely(msg->msg_state == HTTP_MSG_ERROR)) { + if (res->flags & CF_SHUTW) { + /* response errors are most likely due to the + * client aborting the transfer. */ + goto return_cli_abort; + } + goto return_bad_res; + } + DBG_TRACE_LEAVE(STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 1; + } + DBG_TRACE_DEVEL("waiting for the end of the HTTP txn", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + + missing_data_or_waiting: + if (res->flags & CF_SHUTW) + goto return_cli_abort; + + /* stop waiting for data if the input is closed before the end. If the + * client side was already closed, it means that the client has aborted, + * so we don't want to count this as a server abort. Otherwise it's a + * server abort. + */ + if (msg->msg_state < HTTP_MSG_ENDING && res->flags & CF_SHUTR) { + if ((s->req.flags & (CF_SHUTR|CF_SHUTW)) == (CF_SHUTR|CF_SHUTW)) + goto return_cli_abort; + /* If we have some pending data, we continue the processing */ + if (htx_is_empty(htx)) + goto return_srv_abort; + } + + /* When TE: chunked is used, we need to get there again to parse + * remaining chunks even if the server has closed, so we don't want to + * set CF_DONTCLOSE. Similarly when there is a content-leng or if there + * are filters registered on the stream, we don't want to forward a + * close + */ + if ((msg->flags & HTTP_MSGF_XFER_LEN) || HAS_RSP_DATA_FILTERS(s)) + channel_dont_close(res); + + /* We know that more data are expected, but we couldn't send more that + * what we did. So we always set the CF_EXPECT_MORE flag so that the + * system knows it must not set a PUSH on this first part. Interactive + * modes are already handled by the stream sock layer. We must not do + * this in content-length mode because it could present the MSG_MORE + * flag with the last block of forwarded data, which would cause an + * additional delay to be observed by the receiver. + */ + if (HAS_RSP_DATA_FILTERS(s)) + res->flags |= CF_EXPECT_MORE; + + /* the stream handler will take care of timeouts and errors */ + DBG_TRACE_DEVEL("waiting for more data to forward", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA, s, txn); + return 0; + + return_srv_abort: + _HA_ATOMIC_INC(&sess->fe->fe_counters.srv_aborts); + _HA_ATOMIC_INC(&s->be->be_counters.srv_aborts); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->srv_aborts); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.srv_aborts); + stream_inc_http_fail_ctr(s); + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_SRVCL; + goto return_error; + + return_cli_abort: + _HA_ATOMIC_INC(&sess->fe->fe_counters.cli_aborts); + _HA_ATOMIC_INC(&s->be->be_counters.cli_aborts); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->cli_aborts); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.cli_aborts); + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_CLICL; + goto return_error; + + return_int_err: + _HA_ATOMIC_INC(&sess->fe->fe_counters.internal_errors); + _HA_ATOMIC_INC(&s->be->be_counters.internal_errors); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->internal_errors); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.internal_errors); + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_INTERNAL; + goto return_error; + + return_bad_res: + _HA_ATOMIC_INC(&s->be->be_counters.failed_resp); + if (objt_server(s->target)) { + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.failed_resp); + health_adjust(__objt_server(s->target), HANA_STATUS_HTTP_RSP); + } + stream_inc_http_fail_ctr(s); + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_SRVCL; + /* fall through */ + + return_error: + /* don't send any error message as we're in the body */ + http_reply_and_close(s, txn->status, NULL); + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_D; + DBG_TRACE_DEVEL("leaving on error", + STRM_EV_STRM_ANA|STRM_EV_HTTP_ANA|STRM_EV_HTTP_ERR, s, txn); + return 0; +} + +/* Perform an HTTP redirect based on the information in <rule>. The function + * returns zero in case of an irrecoverable error such as too large a request + * to build a valid response, 1 in case of successful redirect (hence the rule + * is final), or 2 if the rule has to be silently skipped. + */ +int http_apply_redirect_rule(struct redirect_rule *rule, struct stream *s, struct http_txn *txn) +{ + struct channel *req = &s->req; + struct channel *res = &s->res; + struct htx *htx; + struct htx_sl *sl; + struct buffer *chunk; + struct ist status, reason, location; + unsigned int flags; + int ret = 1, close = 0; /* Try to keep the connection alive byt default */ + + chunk = alloc_trash_chunk(); + if (!chunk) { + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_RESOURCE; + goto fail; + } + + /* + * Create the location + */ + htx = htxbuf(&req->buf); + switch(rule->type) { + case REDIRECT_TYPE_SCHEME: { + struct http_hdr_ctx ctx; + struct ist path, host; + struct http_uri_parser parser; + + host = ist(""); + ctx.blk = NULL; + if (http_find_header(htx, ist("Host"), &ctx, 0)) + host = ctx.value; + + sl = http_get_stline(htx); + parser = http_uri_parser_init(htx_sl_req_uri(sl)); + path = http_parse_path(&parser); + /* build message using path */ + if (isttest(path)) { + if (rule->flags & REDIRECT_FLAG_DROP_QS) { + int qs = 0; + while (qs < path.len) { + if (*(path.ptr + qs) == '?') { + path.len = qs; + break; + } + qs++; + } + } + } + else + path = ist("/"); + + if (rule->rdr_str) { /* this is an old "redirect" rule */ + /* add scheme */ + if (!chunk_memcat(chunk, rule->rdr_str, rule->rdr_len)) + goto fail; + } + else { + /* add scheme with executing log format */ + chunk->data += build_logline(s, chunk->area + chunk->data, + chunk->size - chunk->data, + &rule->rdr_fmt); + } + /* add "://" + host + path */ + if (!chunk_memcat(chunk, "://", 3) || + !chunk_memcat(chunk, host.ptr, host.len) || + !chunk_memcat(chunk, path.ptr, path.len)) + goto fail; + + /* append a slash at the end of the location if needed and missing */ + if (chunk->data && chunk->area[chunk->data - 1] != '/' && + (rule->flags & REDIRECT_FLAG_APPEND_SLASH)) { + if (chunk->data + 1 >= chunk->size) + goto fail; + chunk->area[chunk->data++] = '/'; + } + break; + } + + case REDIRECT_TYPE_PREFIX: { + struct ist path; + struct http_uri_parser parser; + + sl = http_get_stline(htx); + parser = http_uri_parser_init(htx_sl_req_uri(sl)); + path = http_parse_path(&parser); + /* build message using path */ + if (isttest(path)) { + if (rule->flags & REDIRECT_FLAG_DROP_QS) { + int qs = 0; + while (qs < path.len) { + if (*(path.ptr + qs) == '?') { + path.len = qs; + break; + } + qs++; + } + } + } + else + path = ist("/"); + + if (rule->rdr_str) { /* this is an old "redirect" rule */ + /* add prefix. Note that if prefix == "/", we don't want to + * add anything, otherwise it makes it hard for the user to + * configure a self-redirection. + */ + if (rule->rdr_len != 1 || *rule->rdr_str != '/') { + if (!chunk_memcat(chunk, rule->rdr_str, rule->rdr_len)) + goto fail; + } + } + else { + /* add prefix with executing log format */ + chunk->data += build_logline(s, chunk->area + chunk->data, + chunk->size - chunk->data, + &rule->rdr_fmt); + } + + /* add path */ + if (!chunk_memcat(chunk, path.ptr, path.len)) + goto fail; + + /* append a slash at the end of the location if needed and missing */ + if (chunk->data && chunk->area[chunk->data - 1] != '/' && + (rule->flags & REDIRECT_FLAG_APPEND_SLASH)) { + if (chunk->data + 1 >= chunk->size) + goto fail; + chunk->area[chunk->data++] = '/'; + } + break; + } + case REDIRECT_TYPE_LOCATION: + default: + if (rule->rdr_str) { /* this is an old "redirect" rule */ + /* add location */ + if (!chunk_memcat(chunk, rule->rdr_str, rule->rdr_len)) + goto fail; + } + else { + /* add location with executing log format */ + int len = build_logline(s, chunk->area + chunk->data, + chunk->size - chunk->data, + &rule->rdr_fmt); + if (!len && rule->flags & REDIRECT_FLAG_IGNORE_EMPTY) { + ret = 2; + goto out; + } + + chunk->data += len; + } + break; + } + location = ist2(chunk->area, chunk->data); + + /* + * Create the 30x response + */ + switch (rule->code) { + case 308: + status = ist("308"); + reason = ist("Permanent Redirect"); + break; + case 307: + status = ist("307"); + reason = ist("Temporary Redirect"); + break; + case 303: + status = ist("303"); + reason = ist("See Other"); + break; + case 301: + status = ist("301"); + reason = ist("Moved Permanently"); + break; + case 302: + default: + status = ist("302"); + reason = ist("Found"); + break; + } + + if (!(txn->req.flags & HTTP_MSGF_BODYLESS) && txn->req.msg_state != HTTP_MSG_DONE) + close = 1; + + htx = htx_from_buf(&res->buf); + /* Trim any possible response */ + channel_htx_truncate(&s->res, htx); + flags = (HTX_SL_F_IS_RESP|HTX_SL_F_VER_11|HTX_SL_F_XFER_LEN|HTX_SL_F_BODYLESS); + sl = htx_add_stline(htx, HTX_BLK_RES_SL, flags, ist("HTTP/1.1"), status, reason); + if (!sl) + goto fail; + sl->info.res.status = rule->code; + s->txn->status = rule->code; + + if (close && !htx_add_header(htx, ist("Connection"), ist("close"))) + goto fail; + + if (!htx_add_header(htx, ist("Content-length"), ist("0")) || + !htx_add_header(htx, ist("Location"), location)) + goto fail; + + if (rule->code == 302 || rule->code == 303 || rule->code == 307) { + if (!htx_add_header(htx, ist("Cache-Control"), ist("no-cache"))) + goto fail; + } + + if (rule->cookie_len) { + if (!htx_add_header(htx, ist("Set-Cookie"), ist2(rule->cookie_str, rule->cookie_len))) + goto fail; + } + + if (!htx_add_endof(htx, HTX_BLK_EOH)) + goto fail; + + htx->flags |= HTX_FL_EOM; + htx_to_buf(htx, &res->buf); + if (!http_forward_proxy_resp(s, 1)) + goto fail; + + if (rule->flags & REDIRECT_FLAG_FROM_REQ) { + /* let's log the request time */ + s->logs.tv_request = now; + req->analysers &= AN_REQ_FLT_END; + + if (s->sess->fe == s->be) /* report it if the request was intercepted by the frontend */ + _HA_ATOMIC_INC(&s->sess->fe->fe_counters.intercepted_req); + } + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_LOCAL; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= ((rule->flags & REDIRECT_FLAG_FROM_REQ) ? SF_FINST_R : SF_FINST_H); + + out: + free_trash_chunk(chunk); + return ret; + + fail: + /* If an error occurred, remove the incomplete HTTP response from the + * buffer */ + channel_htx_truncate(res, htxbuf(&res->buf)); + ret = 0; + goto out; +} + +/* This function filters the request header names to only allow [0-9a-zA-Z-] + * characters. Depending on the proxy configuration, headers with a name not + * matching this charset are removed or the request is rejected with a + * 403-Forbidden response if such name are found. It returns HTTP_RULE_RES_CONT + * to continue the request processing or HTTP_RULE_RES_DENY if the request is + * rejected. + */ +static enum rule_result http_req_restrict_header_names(struct stream *s, struct htx *htx, struct proxy *px) +{ + struct htx_blk *blk; + enum rule_result rule_ret = HTTP_RULE_RES_CONT; + + blk = htx_get_first_blk(htx); + while (blk) { + enum htx_blk_type type = htx_get_blk_type(blk); + + if (type == HTX_BLK_HDR) { + struct ist n = htx_get_blk_name(htx, blk); + int i, end = istlen(n); + + for (i = 0; i < end; i++) { + if (!isalnum((unsigned char)n.ptr[i]) && n.ptr[i] != '-') { + break; + } + } + + if (i < end) { + /* Disallowed character found - block the request or remove the header */ + if (px->options2 & PR_O2_RSTRICT_REQ_HDR_NAMES_BLK) + goto block; + blk = htx_remove_blk(htx, blk); + continue; + } + } + if (type == HTX_BLK_EOH) + break; + + blk = htx_get_next_blk(htx, blk); + } + out: + return rule_ret; + block: + /* Block the request returning a 403-Forbidden response */ + s->txn->status = 403; + rule_ret = HTTP_RULE_RES_DENY; + goto out; +} + +/* Replace all headers matching the name <name>. The header value is replaced if + * it matches the regex <re>. <str> is used for the replacement. If <full> is + * set to 1, the full-line is matched and replaced. Otherwise, comma-separated + * values are evaluated one by one. It returns 0 on success and -1 on error. + */ +int http_replace_hdrs(struct stream* s, struct htx *htx, struct ist name, + const char *str, struct my_regex *re, int full) +{ + struct http_hdr_ctx ctx; + struct buffer *output = get_trash_chunk(); + + ctx.blk = NULL; + while (http_find_header(htx, name, &ctx, full)) { + if (!regex_exec_match2(re, ctx.value.ptr, ctx.value.len, MAX_MATCH, pmatch, 0)) + continue; + + output->data = exp_replace(output->area, output->size, ctx.value.ptr, str, pmatch); + if (output->data == -1) + return -1; + if (!http_replace_header_value(htx, &ctx, ist2(output->area, output->data))) + return -1; + } + return 0; +} + +/* This function executes one of the set-{method,path,query,uri} actions. It + * takes the string from the variable 'replace' with length 'len', then modifies + * the relevant part of the request line accordingly. Then it updates various + * pointers to the next elements which were moved, and the total buffer length. + * It finds the action to be performed in p[2], previously filled by function + * parse_set_req_line(). It returns 0 in case of success, -1 in case of internal + * error, though this can be revisited when this code is finally exploited. + * + * 'action' can be '0' to replace method, '1' to replace path, '2' to replace + * query string, 3 to replace uri or 4 to replace the path+query. + * + * In query string case, the mark question '?' must be set at the start of the + * string by the caller, event if the replacement query string is empty. + */ +int http_req_replace_stline(int action, const char *replace, int len, + struct proxy *px, struct stream *s) +{ + struct htx *htx = htxbuf(&s->req.buf); + + switch (action) { + case 0: // method + if (!http_replace_req_meth(htx, ist2(replace, len))) + return -1; + break; + + case 1: // path + if (!http_replace_req_path(htx, ist2(replace, len), 0)) + return -1; + break; + + case 2: // query + if (!http_replace_req_query(htx, ist2(replace, len))) + return -1; + break; + + case 3: // uri + if (!http_replace_req_uri(htx, ist2(replace, len))) + return -1; + break; + + case 4: // path + query + if (!http_replace_req_path(htx, ist2(replace, len), 1)) + return -1; + break; + + default: + return -1; + } + return 0; +} + +/* This function replace the HTTP status code and the associated message. The + * variable <status> contains the new status code. This function never fails. It + * returns 0 in case of success, -1 in case of internal error. + */ +int http_res_set_status(unsigned int status, struct ist reason, struct stream *s) +{ + struct htx *htx = htxbuf(&s->res.buf); + char *res; + + chunk_reset(&trash); + res = ultoa_o(status, trash.area, trash.size); + trash.data = res - trash.area; + + /* Do we have a custom reason format string? */ + if (!isttest(reason)) { + const char *str = http_get_reason(status); + reason = ist(str); + } + + if (!http_replace_res_status(htx, ist2(trash.area, trash.data), reason)) + return -1; + s->txn->status = status; + return 0; +} + +/* Executes the http-request rules <rules> for stream <s>, proxy <px> and + * transaction <txn>. Returns the verdict of the first rule that prevents + * further processing of the request (auth, deny, ...), and defaults to + * HTTP_RULE_RES_STOP if it executed all rules or stopped on an allow, or + * HTTP_RULE_RES_CONT if the last rule was reached. It may set the TX_CLTARPIT + * on txn->flags if it encounters a tarpit rule. If <deny_status> is not NULL + * and a deny/tarpit rule is matched, it will be filled with this rule's deny + * status. + */ +static enum rule_result http_req_get_intercept_rule(struct proxy *px, struct list *def_rules, + struct list *rules, struct stream *s) +{ + struct session *sess = strm_sess(s); + struct http_txn *txn = s->txn; + struct act_rule *rule; + enum rule_result rule_ret = HTTP_RULE_RES_CONT; + int act_opts = 0; + + /* If "the current_rule_list" match the executed rule list, we are in + * resume condition. If a resume is needed it is always in the action + * and never in the ACL or converters. In this case, we initialise the + * current rule, and go to the action execution point. + */ + if (s->current_rule) { + rule = s->current_rule; + s->current_rule = NULL; + if (s->current_rule_list == rules || (def_rules && s->current_rule_list == def_rules)) + goto resume_execution; + } + s->current_rule_list = ((!def_rules || s->current_rule_list == def_rules) ? rules : def_rules); + + restart: + /* start the ruleset evaluation in strict mode */ + txn->req.flags &= ~HTTP_MSGF_SOFT_RW; + + list_for_each_entry(rule, s->current_rule_list, list) { + /* check optional condition */ + if (rule->cond) { + int ret; + + ret = acl_exec_cond(rule->cond, px, sess, s, SMP_OPT_DIR_REQ|SMP_OPT_FINAL); + ret = acl_pass(ret); + + if (rule->cond->pol == ACL_COND_UNLESS) + ret = !ret; + + if (!ret) /* condition not matched */ + continue; + } + + act_opts |= ACT_OPT_FIRST; + resume_execution: + if (rule->kw->flags & KWF_EXPERIMENTAL) + mark_tainted(TAINTED_ACTION_EXP_EXECUTED); + + /* Always call the action function if defined */ + if (rule->action_ptr) { + if ((s->req.flags & CF_READ_ERROR) || + ((s->req.flags & (CF_SHUTR|CF_READ_NULL)) && + (px->options & PR_O_ABRT_CLOSE))) + act_opts |= ACT_OPT_FINAL; + + switch (rule->action_ptr(rule, px, sess, s, act_opts)) { + case ACT_RET_CONT: + break; + case ACT_RET_STOP: + rule_ret = HTTP_RULE_RES_STOP; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_YIELD: + s->current_rule = rule; + rule_ret = HTTP_RULE_RES_YIELD; + goto end; + case ACT_RET_ERR: + rule_ret = HTTP_RULE_RES_ERROR; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_DONE: + rule_ret = HTTP_RULE_RES_DONE; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_DENY: + if (txn->status == -1) + txn->status = 403; + rule_ret = HTTP_RULE_RES_DENY; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_ABRT: + rule_ret = HTTP_RULE_RES_ABRT; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_INV: + rule_ret = HTTP_RULE_RES_BADREQ; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + } + continue; /* eval the next rule */ + } + + /* If not action function defined, check for known actions */ + switch (rule->action) { + case ACT_ACTION_ALLOW: + rule_ret = HTTP_RULE_RES_STOP; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + + case ACT_ACTION_DENY: + txn->status = rule->arg.http_reply->status; + txn->http_reply = rule->arg.http_reply; + rule_ret = HTTP_RULE_RES_DENY; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + + case ACT_HTTP_REQ_TARPIT: + txn->flags |= TX_CLTARPIT; + txn->status = rule->arg.http_reply->status; + txn->http_reply = rule->arg.http_reply; + rule_ret = HTTP_RULE_RES_DENY; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + + case ACT_HTTP_REDIR: { + int ret = http_apply_redirect_rule(rule->arg.redir, s, txn); + + if (ret == 2) // 2 == skip + break; + + rule_ret = ret ? HTTP_RULE_RES_ABRT : HTTP_RULE_RES_ERROR; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + } + + /* other flags exists, but normally, they never be matched. */ + default: + break; + } + } + + if (def_rules && s->current_rule_list == def_rules) { + s->current_rule_list = rules; + goto restart; + } + + end: + /* if the ruleset evaluation is finished reset the strict mode */ + if (rule_ret != HTTP_RULE_RES_YIELD) + txn->req.flags &= ~HTTP_MSGF_SOFT_RW; + + /* we reached the end of the rules, nothing to report */ + return rule_ret; +} + +/* Executes the http-response rules <rules> for stream <s> and proxy <px>. It + * returns one of 5 possible statuses: HTTP_RULE_RES_CONT, HTTP_RULE_RES_STOP, + * HTTP_RULE_RES_DONE, HTTP_RULE_RES_YIELD, or HTTP_RULE_RES_BADREQ. If *CONT + * is returned, the process can continue the evaluation of next rule list. If + * *STOP or *DONE is returned, the process must stop the evaluation. If *BADREQ + * is returned, it means the operation could not be processed and a server error + * must be returned. If *YIELD is returned, the caller must call again the + * function with the same context. + */ +static enum rule_result http_res_get_intercept_rule(struct proxy *px, struct list *def_rules, + struct list *rules, struct stream *s) +{ + struct session *sess = strm_sess(s); + struct http_txn *txn = s->txn; + struct act_rule *rule; + enum rule_result rule_ret = HTTP_RULE_RES_CONT; + int act_opts = 0; + + /* If "the current_rule_list" match the executed rule list, we are in + * resume condition. If a resume is needed it is always in the action + * and never in the ACL or converters. In this case, we initialise the + * current rule, and go to the action execution point. + */ + if (s->current_rule) { + rule = s->current_rule; + s->current_rule = NULL; + if (s->current_rule_list == rules || (def_rules && s->current_rule_list == def_rules)) + goto resume_execution; + } + s->current_rule_list = ((!def_rules || s->current_rule_list == def_rules) ? rules : def_rules); + + restart: + + /* start the ruleset evaluation in strict mode */ + txn->rsp.flags &= ~HTTP_MSGF_SOFT_RW; + + list_for_each_entry(rule, s->current_rule_list, list) { + /* check optional condition */ + if (rule->cond) { + int ret; + + ret = acl_exec_cond(rule->cond, px, sess, s, SMP_OPT_DIR_RES|SMP_OPT_FINAL); + ret = acl_pass(ret); + + if (rule->cond->pol == ACL_COND_UNLESS) + ret = !ret; + + if (!ret) /* condition not matched */ + continue; + } + + act_opts |= ACT_OPT_FIRST; +resume_execution: + if (rule->kw->flags & KWF_EXPERIMENTAL) + mark_tainted(TAINTED_ACTION_EXP_EXECUTED); + + /* Always call the action function if defined */ + if (rule->action_ptr) { + if ((s->req.flags & CF_READ_ERROR) || + ((s->req.flags & (CF_SHUTR|CF_READ_NULL)) && + (px->options & PR_O_ABRT_CLOSE))) + act_opts |= ACT_OPT_FINAL; + + switch (rule->action_ptr(rule, px, sess, s, act_opts)) { + case ACT_RET_CONT: + break; + case ACT_RET_STOP: + rule_ret = HTTP_RULE_RES_STOP; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_YIELD: + s->current_rule = rule; + rule_ret = HTTP_RULE_RES_YIELD; + goto end; + case ACT_RET_ERR: + rule_ret = HTTP_RULE_RES_ERROR; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_DONE: + rule_ret = HTTP_RULE_RES_DONE; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_DENY: + if (txn->status == -1) + txn->status = 502; + rule_ret = HTTP_RULE_RES_DENY; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_ABRT: + rule_ret = HTTP_RULE_RES_ABRT; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + case ACT_RET_INV: + rule_ret = HTTP_RULE_RES_BADREQ; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + } + continue; /* eval the next rule */ + } + + /* If not action function defined, check for known actions */ + switch (rule->action) { + case ACT_ACTION_ALLOW: + rule_ret = HTTP_RULE_RES_STOP; /* "allow" rules are OK */ + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + + case ACT_ACTION_DENY: + txn->status = rule->arg.http_reply->status; + txn->http_reply = rule->arg.http_reply; + rule_ret = HTTP_RULE_RES_DENY; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + + case ACT_HTTP_REDIR: { + int ret = http_apply_redirect_rule(rule->arg.redir, s, txn); + + if (ret == 2) // 2 == skip + break; + + rule_ret = ret ? HTTP_RULE_RES_ABRT : HTTP_RULE_RES_ERROR; + s->last_rule_file = rule->conf.file; + s->last_rule_line = rule->conf.line; + goto end; + } + /* other flags exists, but normally, they never be matched. */ + default: + break; + } + } + + if (def_rules && s->current_rule_list == def_rules) { + s->current_rule_list = rules; + goto restart; + } + + end: + /* if the ruleset evaluation is finished reset the strict mode */ + if (rule_ret != HTTP_RULE_RES_YIELD) + txn->rsp.flags &= ~HTTP_MSGF_SOFT_RW; + + /* we reached the end of the rules, nothing to report */ + return rule_ret; +} + +/* Executes backend and frontend http-after-response rules for the stream <s>, + * in that order. it return 1 on success and 0 on error. It is the caller + * responsibility to catch error or ignore it. If it catches it, this function + * may be called a second time, for the internal error. + */ +int http_eval_after_res_rules(struct stream *s) +{ + struct list *def_rules, *rules; + struct session *sess = s->sess; + enum rule_result ret = HTTP_RULE_RES_CONT; + + /* Eval after-response ruleset only if the reply is not const */ + if (s->txn->flags & TX_CONST_REPLY) + goto end; + + /* prune the request variables if not already done and swap to the response variables. */ + if (s->vars_reqres.scope != SCOPE_RES) { + if (!LIST_ISEMPTY(&s->vars_reqres.head)) + vars_prune(&s->vars_reqres, s->sess, s); + vars_init_head(&s->vars_reqres, SCOPE_RES); + } + + def_rules = (s->be->defpx ? &s->be->defpx->http_after_res_rules : NULL); + rules = &s->be->http_after_res_rules; + + ret = http_res_get_intercept_rule(s->be, def_rules, rules, s); + if ((ret == HTTP_RULE_RES_CONT || ret == HTTP_RULE_RES_STOP) && sess->fe != s->be) { + def_rules = ((sess->fe->defpx && sess->fe->defpx != s->be->defpx) ? &sess->fe->defpx->http_after_res_rules : NULL); + rules = &sess->fe->http_after_res_rules; + ret = http_res_get_intercept_rule(sess->fe, def_rules, rules, s); + } + + end: + /* All other codes than CONTINUE, STOP or DONE are forbidden */ + return (ret == HTTP_RULE_RES_CONT || ret == HTTP_RULE_RES_STOP || ret == HTTP_RULE_RES_DONE); +} + +/* + * Manage client-side cookie. It can impact performance by about 2% so it is + * desirable to call it only when needed. This code is quite complex because + * of the multiple very crappy and ambiguous syntaxes we have to support. it + * highly recommended not to touch this part without a good reason ! + */ +static void http_manage_client_side_cookies(struct stream *s, struct channel *req) +{ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct htx *htx; + struct http_hdr_ctx ctx; + char *hdr_beg, *hdr_end, *del_from; + char *prev, *att_beg, *att_end, *equal, *val_beg, *val_end, *next; + int preserve_hdr; + + htx = htxbuf(&req->buf); + ctx.blk = NULL; + while (http_find_header(htx, ist("Cookie"), &ctx, 1)) { + int is_first = 1; + del_from = NULL; /* nothing to be deleted */ + preserve_hdr = 0; /* assume we may kill the whole header */ + + /* Now look for cookies. Conforming to RFC2109, we have to support + * attributes whose name begin with a '$', and associate them with + * the right cookie, if we want to delete this cookie. + * So there are 3 cases for each cookie read : + * 1) it's a special attribute, beginning with a '$' : ignore it. + * 2) it's a server id cookie that we *MAY* want to delete : save + * some pointers on it (last semi-colon, beginning of cookie...) + * 3) it's an application cookie : we *MAY* have to delete a previous + * "special" cookie. + * At the end of loop, if a "special" cookie remains, we may have to + * remove it. If no application cookie persists in the header, we + * *MUST* delete it. + * + * Note: RFC2965 is unclear about the processing of spaces around + * the equal sign in the ATTR=VALUE form. A careful inspection of + * the RFC explicitly allows spaces before it, and not within the + * tokens (attrs or values). An inspection of RFC2109 allows that + * too but section 10.1.3 lets one think that spaces may be allowed + * after the equal sign too, resulting in some (rare) buggy + * implementations trying to do that. So let's do what servers do. + * Latest ietf draft forbids spaces all around. Also, earlier RFCs + * allowed quoted strings in values, with any possible character + * after a backslash, including control chars and delimiters, which + * causes parsing to become ambiguous. Browsers also allow spaces + * within values even without quotes. + * + * We have to keep multiple pointers in order to support cookie + * removal at the beginning, middle or end of header without + * corrupting the header. All of these headers are valid : + * + * hdr_beg hdr_end + * | | + * v | + * NAME1=VALUE1;NAME2=VALUE2;NAME3=VALUE3 | + * NAME1=VALUE1;NAME2_ONLY ;NAME3=VALUE3 v + * NAME1 = VALUE 1 ; NAME2 = VALUE2 ; NAME3 = VALUE3 + * | | | | | | | + * | | | | | | | + * | | | | | | +--> next + * | | | | | +----> val_end + * | | | | +-----------> val_beg + * | | | +--------------> equal + * | | +----------------> att_end + * | +---------------------> att_beg + * +--------------------------> prev + * + */ + hdr_beg = ctx.value.ptr; + hdr_end = hdr_beg + ctx.value.len; + for (prev = hdr_beg; prev < hdr_end; prev = next) { + /* Iterate through all cookies on this line */ + + /* find att_beg */ + att_beg = prev; + if (!is_first) + att_beg++; + is_first = 0; + + while (att_beg < hdr_end && HTTP_IS_SPHT(*att_beg)) + att_beg++; + + /* find att_end : this is the first character after the last non + * space before the equal. It may be equal to hdr_end. + */ + equal = att_end = att_beg; + while (equal < hdr_end) { + if (*equal == '=' || *equal == ',' || *equal == ';') + break; + if (HTTP_IS_SPHT(*equal++)) + continue; + att_end = equal; + } + + /* here, <equal> points to '=', a delimiter or the end. <att_end> + * is between <att_beg> and <equal>, both may be identical. + */ + /* look for end of cookie if there is an equal sign */ + if (equal < hdr_end && *equal == '=') { + /* look for the beginning of the value */ + val_beg = equal + 1; + while (val_beg < hdr_end && HTTP_IS_SPHT(*val_beg)) + val_beg++; + + /* find the end of the value, respecting quotes */ + next = http_find_cookie_value_end(val_beg, hdr_end); + + /* make val_end point to the first white space or delimiter after the value */ + val_end = next; + while (val_end > val_beg && HTTP_IS_SPHT(*(val_end - 1))) + val_end--; + } + else + val_beg = val_end = next = equal; + + /* We have nothing to do with attributes beginning with + * '$'. However, they will automatically be removed if a + * header before them is removed, since they're supposed + * to be linked together. + */ + if (*att_beg == '$') + continue; + + /* Ignore cookies with no equal sign */ + if (equal == next) { + /* This is not our cookie, so we must preserve it. But if we already + * scheduled another cookie for removal, we cannot remove the + * complete header, but we can remove the previous block itself. + */ + preserve_hdr = 1; + if (del_from != NULL) { + int delta = http_del_hdr_value(hdr_beg, hdr_end, &del_from, prev); + val_end += delta; + next += delta; + hdr_end += delta; + prev = del_from; + del_from = NULL; + } + continue; + } + + /* if there are spaces around the equal sign, we need to + * strip them otherwise we'll get trouble for cookie captures, + * or even for rewrites. Since this happens extremely rarely, + * it does not hurt performance. + */ + if (unlikely(att_end != equal || val_beg > equal + 1)) { + int stripped_before = 0; + int stripped_after = 0; + + if (att_end != equal) { + memmove(att_end, equal, hdr_end - equal); + stripped_before = (att_end - equal); + equal += stripped_before; + val_beg += stripped_before; + } + + if (val_beg > equal + 1) { + memmove(equal + 1, val_beg, hdr_end + stripped_before - val_beg); + stripped_after = (equal + 1) - val_beg; + val_beg += stripped_after; + stripped_before += stripped_after; + } + + val_end += stripped_before; + next += stripped_before; + hdr_end += stripped_before; + } + /* now everything is as on the diagram above */ + + /* First, let's see if we want to capture this cookie. We check + * that we don't already have a client side cookie, because we + * can only capture one. Also as an optimisation, we ignore + * cookies shorter than the declared name. + */ + if (sess->fe->capture_name != NULL && txn->cli_cookie == NULL && + (val_end - att_beg >= sess->fe->capture_namelen) && + memcmp(att_beg, sess->fe->capture_name, sess->fe->capture_namelen) == 0) { + int log_len = val_end - att_beg; + + if ((txn->cli_cookie = pool_alloc(pool_head_capture)) == NULL) { + ha_alert("HTTP logging : out of memory.\n"); + } else { + if (log_len > sess->fe->capture_len) + log_len = sess->fe->capture_len; + memcpy(txn->cli_cookie, att_beg, log_len); + txn->cli_cookie[log_len] = 0; + } + } + + /* Persistence cookies in passive, rewrite or insert mode have the + * following form : + * + * Cookie: NAME=SRV[|<lastseen>[|<firstseen>]] + * + * For cookies in prefix mode, the form is : + * + * Cookie: NAME=SRV~VALUE + */ + if ((att_end - att_beg == s->be->cookie_len) && (s->be->cookie_name != NULL) && + (memcmp(att_beg, s->be->cookie_name, att_end - att_beg) == 0)) { + struct server *srv = s->be->srv; + char *delim; + + /* if we're in cookie prefix mode, we'll search the delimiter so that we + * have the server ID between val_beg and delim, and the original cookie between + * delim+1 and val_end. Otherwise, delim==val_end : + * + * hdr_beg + * | + * v + * NAME=SRV; # in all but prefix modes + * NAME=SRV~OPAQUE ; # in prefix mode + * || || | |+-> next + * || || | +--> val_end + * || || +---------> delim + * || |+------------> val_beg + * || +-------------> att_end = equal + * |+-----------------> att_beg + * +------------------> prev + * + */ + if (s->be->ck_opts & PR_CK_PFX) { + for (delim = val_beg; delim < val_end; delim++) + if (*delim == COOKIE_DELIM) + break; + } + else { + char *vbar1; + delim = val_end; + /* Now check if the cookie contains a date field, which would + * appear after a vertical bar ('|') just after the server name + * and before the delimiter. + */ + vbar1 = memchr(val_beg, COOKIE_DELIM_DATE, val_end - val_beg); + if (vbar1) { + /* OK, so left of the bar is the server's cookie and + * right is the last seen date. It is a base64 encoded + * 30-bit value representing the UNIX date since the + * epoch in 4-second quantities. + */ + int val; + delim = vbar1++; + if (val_end - vbar1 >= 5) { + val = b64tos30(vbar1); + if (val > 0) + txn->cookie_last_date = val << 2; + } + /* look for a second vertical bar */ + vbar1 = memchr(vbar1, COOKIE_DELIM_DATE, val_end - vbar1); + if (vbar1 && (val_end - vbar1 > 5)) { + val = b64tos30(vbar1 + 1); + if (val > 0) + txn->cookie_first_date = val << 2; + } + } + } + + /* if the cookie has an expiration date and the proxy wants to check + * it, then we do that now. We first check if the cookie is too old, + * then only if it has expired. We detect strict overflow because the + * time resolution here is not great (4 seconds). Cookies with dates + * in the future are ignored if their offset is beyond one day. This + * allows an admin to fix timezone issues without expiring everyone + * and at the same time avoids keeping unwanted side effects for too + * long. + */ + if (txn->cookie_first_date && s->be->cookie_maxlife && + (((signed)(date.tv_sec - txn->cookie_first_date) > (signed)s->be->cookie_maxlife) || + ((signed)(txn->cookie_first_date - date.tv_sec) > 86400))) { + txn->flags &= ~TX_CK_MASK; + txn->flags |= TX_CK_OLD; + delim = val_beg; // let's pretend we have not found the cookie + txn->cookie_first_date = 0; + txn->cookie_last_date = 0; + } + else if (txn->cookie_last_date && s->be->cookie_maxidle && + (((signed)(date.tv_sec - txn->cookie_last_date) > (signed)s->be->cookie_maxidle) || + ((signed)(txn->cookie_last_date - date.tv_sec) > 86400))) { + txn->flags &= ~TX_CK_MASK; + txn->flags |= TX_CK_EXPIRED; + delim = val_beg; // let's pretend we have not found the cookie + txn->cookie_first_date = 0; + txn->cookie_last_date = 0; + } + + /* Here, we'll look for the first running server which supports the cookie. + * This allows to share a same cookie between several servers, for example + * to dedicate backup servers to specific servers only. + * However, to prevent clients from sticking to cookie-less backup server + * when they have incidentely learned an empty cookie, we simply ignore + * empty cookies and mark them as invalid. + * The same behaviour is applied when persistence must be ignored. + */ + if ((delim == val_beg) || (s->flags & (SF_IGNORE_PRST | SF_ASSIGNED))) + srv = NULL; + + while (srv) { + if (srv->cookie && (srv->cklen == delim - val_beg) && + !memcmp(val_beg, srv->cookie, delim - val_beg)) { + if ((srv->cur_state != SRV_ST_STOPPED) || + (s->be->options & PR_O_PERSIST) || + (s->flags & SF_FORCE_PRST)) { + /* we found the server and we can use it */ + txn->flags &= ~TX_CK_MASK; + txn->flags |= (srv->cur_state != SRV_ST_STOPPED) ? TX_CK_VALID : TX_CK_DOWN; + s->flags |= SF_DIRECT | SF_ASSIGNED; + s->target = &srv->obj_type; + break; + } else { + /* we found a server, but it's down, + * mark it as such and go on in case + * another one is available. + */ + txn->flags &= ~TX_CK_MASK; + txn->flags |= TX_CK_DOWN; + } + } + srv = srv->next; + } + + if (!srv && !(txn->flags & (TX_CK_DOWN|TX_CK_EXPIRED|TX_CK_OLD))) { + /* no server matched this cookie or we deliberately skipped it */ + txn->flags &= ~TX_CK_MASK; + if ((s->flags & (SF_IGNORE_PRST | SF_ASSIGNED))) + txn->flags |= TX_CK_UNUSED; + else + txn->flags |= TX_CK_INVALID; + } + + /* depending on the cookie mode, we may have to either : + * - delete the complete cookie if we're in insert+indirect mode, so that + * the server never sees it ; + * - remove the server id from the cookie value, and tag the cookie as an + * application cookie so that it does not get accidentally removed later, + * if we're in cookie prefix mode + */ + if ((s->be->ck_opts & PR_CK_PFX) && (delim != val_end)) { + int delta; /* negative */ + + memmove(val_beg, delim + 1, hdr_end - (delim + 1)); + delta = val_beg - (delim + 1); + val_end += delta; + next += delta; + hdr_end += delta; + del_from = NULL; + preserve_hdr = 1; /* we want to keep this cookie */ + } + else if (del_from == NULL && + (s->be->ck_opts & (PR_CK_INS | PR_CK_IND)) == (PR_CK_INS | PR_CK_IND)) { + del_from = prev; + } + } + else { + /* This is not our cookie, so we must preserve it. But if we already + * scheduled another cookie for removal, we cannot remove the + * complete header, but we can remove the previous block itself. + */ + preserve_hdr = 1; + + if (del_from != NULL) { + int delta = http_del_hdr_value(hdr_beg, hdr_end, &del_from, prev); + if (att_beg >= del_from) + att_beg += delta; + if (att_end >= del_from) + att_end += delta; + val_beg += delta; + val_end += delta; + next += delta; + hdr_end += delta; + prev = del_from; + del_from = NULL; + } + } + + } /* for each cookie */ + + + /* There are no more cookies on this line. + * We may still have one (or several) marked for deletion at the + * end of the line. We must do this now in two ways : + * - if some cookies must be preserved, we only delete from the + * mark to the end of line ; + * - if nothing needs to be preserved, simply delete the whole header + */ + if (del_from) { + hdr_end = (preserve_hdr ? del_from : hdr_beg); + } + if ((hdr_end - hdr_beg) != ctx.value.len) { + if (hdr_beg != hdr_end) + htx_change_blk_value_len(htx, ctx.blk, hdr_end - hdr_beg); + else + http_remove_header(htx, &ctx); + } + } /* for each "Cookie header */ +} + +/* + * Manage server-side cookies. It can impact performance by about 2% so it is + * desirable to call it only when needed. This function is also used when we + * just need to know if there is a cookie (eg: for check-cache). + */ +static void http_manage_server_side_cookies(struct stream *s, struct channel *res) +{ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct htx *htx; + struct http_hdr_ctx ctx; + struct server *srv; + char *hdr_beg, *hdr_end; + char *prev, *att_beg, *att_end, *equal, *val_beg, *val_end, *next; + int is_cookie2 = 0; + + htx = htxbuf(&res->buf); + + ctx.blk = NULL; + while (1) { + int is_first = 1; + + if (is_cookie2 || !http_find_header(htx, ist("Set-Cookie"), &ctx, 1)) { + if (!http_find_header(htx, ist("Set-Cookie2"), &ctx, 1)) + break; + is_cookie2 = 1; + } + + /* OK, right now we know we have a Set-Cookie* at hdr_beg, and + * <prev> points to the colon. + */ + txn->flags |= TX_SCK_PRESENT; + + /* Maybe we only wanted to see if there was a Set-Cookie (eg: + * check-cache is enabled) and we are not interested in checking + * them. Warning, the cookie capture is declared in the frontend. + */ + if (s->be->cookie_name == NULL && sess->fe->capture_name == NULL) + break; + + /* OK so now we know we have to process this response cookie. + * The format of the Set-Cookie header is slightly different + * from the format of the Cookie header in that it does not + * support the comma as a cookie delimiter (thus the header + * cannot be folded) because the Expires attribute described in + * the original Netscape's spec may contain an unquoted date + * with a comma inside. We have to live with this because + * many browsers don't support Max-Age and some browsers don't + * support quoted strings. However the Set-Cookie2 header is + * clean. + * + * We have to keep multiple pointers in order to support cookie + * removal at the beginning, middle or end of header without + * corrupting the header (in case of set-cookie2). A special + * pointer, <scav> points to the beginning of the set-cookie-av + * fields after the first semi-colon. The <next> pointer points + * either to the end of line (set-cookie) or next unquoted comma + * (set-cookie2). All of these headers are valid : + * + * hdr_beg hdr_end + * | | + * v | + * NAME1 = VALUE 1 ; Secure; Path="/" | + * NAME=VALUE; Secure; Expires=Thu, 01-Jan-1970 00:00:01 GMT v + * NAME = VALUE ; Secure; Expires=Thu, 01-Jan-1970 00:00:01 GMT + * NAME1 = VALUE 1 ; Max-Age=0, NAME2=VALUE2; Discard + * | | | | | | | | + * | | | | | | | +-> next + * | | | | | | +------------> scav + * | | | | | +--------------> val_end + * | | | | +--------------------> val_beg + * | | | +----------------------> equal + * | | +------------------------> att_end + * | +----------------------------> att_beg + * +------------------------------> prev + * -------------------------------> hdr_beg + */ + hdr_beg = ctx.value.ptr; + hdr_end = hdr_beg + ctx.value.len; + for (prev = hdr_beg; prev < hdr_end; prev = next) { + + /* Iterate through all cookies on this line */ + + /* find att_beg */ + att_beg = prev; + if (!is_first) + att_beg++; + is_first = 0; + + while (att_beg < hdr_end && HTTP_IS_SPHT(*att_beg)) + att_beg++; + + /* find att_end : this is the first character after the last non + * space before the equal. It may be equal to hdr_end. + */ + equal = att_end = att_beg; + + while (equal < hdr_end) { + if (*equal == '=' || *equal == ';' || (is_cookie2 && *equal == ',')) + break; + if (HTTP_IS_SPHT(*equal++)) + continue; + att_end = equal; + } + + /* here, <equal> points to '=', a delimiter or the end. <att_end> + * is between <att_beg> and <equal>, both may be identical. + */ + + /* look for end of cookie if there is an equal sign */ + if (equal < hdr_end && *equal == '=') { + /* look for the beginning of the value */ + val_beg = equal + 1; + while (val_beg < hdr_end && HTTP_IS_SPHT(*val_beg)) + val_beg++; + + /* find the end of the value, respecting quotes */ + next = http_find_cookie_value_end(val_beg, hdr_end); + + /* make val_end point to the first white space or delimiter after the value */ + val_end = next; + while (val_end > val_beg && HTTP_IS_SPHT(*(val_end - 1))) + val_end--; + } + else { + /* <equal> points to next comma, semi-colon or EOL */ + val_beg = val_end = next = equal; + } + + if (next < hdr_end) { + /* Set-Cookie2 supports multiple cookies, and <next> points to + * a colon or semi-colon before the end. So skip all attr-value + * pairs and look for the next comma. For Set-Cookie, since + * commas are permitted in values, skip to the end. + */ + if (is_cookie2) + next = http_find_hdr_value_end(next, hdr_end); + else + next = hdr_end; + } + + /* Now everything is as on the diagram above */ + + /* Ignore cookies with no equal sign */ + if (equal == val_end) + continue; + + /* If there are spaces around the equal sign, we need to + * strip them otherwise we'll get trouble for cookie captures, + * or even for rewrites. Since this happens extremely rarely, + * it does not hurt performance. + */ + if (unlikely(att_end != equal || val_beg > equal + 1)) { + int stripped_before = 0; + int stripped_after = 0; + + if (att_end != equal) { + memmove(att_end, equal, hdr_end - equal); + stripped_before = (att_end - equal); + equal += stripped_before; + val_beg += stripped_before; + } + + if (val_beg > equal + 1) { + memmove(equal + 1, val_beg, hdr_end + stripped_before - val_beg); + stripped_after = (equal + 1) - val_beg; + val_beg += stripped_after; + stripped_before += stripped_after; + } + + val_end += stripped_before; + next += stripped_before; + hdr_end += stripped_before; + + htx_change_blk_value_len(htx, ctx.blk, hdr_end - hdr_beg); + ctx.value.len = hdr_end - hdr_beg; + } + + /* First, let's see if we want to capture this cookie. We check + * that we don't already have a server side cookie, because we + * can only capture one. Also as an optimisation, we ignore + * cookies shorter than the declared name. + */ + if (sess->fe->capture_name != NULL && + txn->srv_cookie == NULL && + (val_end - att_beg >= sess->fe->capture_namelen) && + memcmp(att_beg, sess->fe->capture_name, sess->fe->capture_namelen) == 0) { + int log_len = val_end - att_beg; + if ((txn->srv_cookie = pool_alloc(pool_head_capture)) == NULL) { + ha_alert("HTTP logging : out of memory.\n"); + } + else { + if (log_len > sess->fe->capture_len) + log_len = sess->fe->capture_len; + memcpy(txn->srv_cookie, att_beg, log_len); + txn->srv_cookie[log_len] = 0; + } + } + + srv = objt_server(s->target); + /* now check if we need to process it for persistence */ + if (!(s->flags & SF_IGNORE_PRST) && + (att_end - att_beg == s->be->cookie_len) && (s->be->cookie_name != NULL) && + (memcmp(att_beg, s->be->cookie_name, att_end - att_beg) == 0)) { + /* assume passive cookie by default */ + txn->flags &= ~TX_SCK_MASK; + txn->flags |= TX_SCK_FOUND; + + /* If the cookie is in insert mode on a known server, we'll delete + * this occurrence because we'll insert another one later. + * We'll delete it too if the "indirect" option is set and we're in + * a direct access. + */ + if (s->be->ck_opts & PR_CK_PSV) { + /* The "preserve" flag was set, we don't want to touch the + * server's cookie. + */ + } + else if ((srv && (s->be->ck_opts & PR_CK_INS)) || + ((s->flags & SF_DIRECT) && (s->be->ck_opts & PR_CK_IND))) { + /* this cookie must be deleted */ + if (prev == hdr_beg && next == hdr_end) { + /* whole header */ + http_remove_header(htx, &ctx); + /* note: while both invalid now, <next> and <hdr_end> + * are still equal, so the for() will stop as expected. + */ + } else { + /* just remove the value */ + int delta = http_del_hdr_value(hdr_beg, hdr_end, &prev, next); + next = prev; + hdr_end += delta; + } + txn->flags &= ~TX_SCK_MASK; + txn->flags |= TX_SCK_DELETED; + /* and go on with next cookie */ + } + else if (srv && srv->cookie && (s->be->ck_opts & PR_CK_RW)) { + /* replace bytes val_beg->val_end with the cookie name associated + * with this server since we know it. + */ + int sliding, delta; + + ctx.value = ist2(val_beg, val_end - val_beg); + ctx.lws_before = ctx.lws_after = 0; + http_replace_header_value(htx, &ctx, ist2(srv->cookie, srv->cklen)); + delta = srv->cklen - (val_end - val_beg); + sliding = (ctx.value.ptr - val_beg); + hdr_beg += sliding; + val_beg += sliding; + next += sliding + delta; + hdr_end += sliding + delta; + + txn->flags &= ~TX_SCK_MASK; + txn->flags |= TX_SCK_REPLACED; + } + else if (srv && srv->cookie && (s->be->ck_opts & PR_CK_PFX)) { + /* insert the cookie name associated with this server + * before existing cookie, and insert a delimiter between them.. + */ + int sliding, delta; + ctx.value = ist2(val_beg, 0); + ctx.lws_before = ctx.lws_after = 0; + http_replace_header_value(htx, &ctx, ist2(srv->cookie, srv->cklen + 1)); + delta = srv->cklen + 1; + sliding = (ctx.value.ptr - val_beg); + hdr_beg += sliding; + val_beg += sliding; + next += sliding + delta; + hdr_end += sliding + delta; + + val_beg[srv->cklen] = COOKIE_DELIM; + txn->flags &= ~TX_SCK_MASK; + txn->flags |= TX_SCK_REPLACED; + } + } + /* that's done for this cookie, check the next one on the same + * line when next != hdr_end (only if is_cookie2). + */ + } + } +} + +/* + * Parses the Cache-Control and Pragma request header fields to determine if + * the request may be served from the cache and/or if it is cacheable. Updates + * s->txn->flags. + */ +void http_check_request_for_cacheability(struct stream *s, struct channel *req) +{ + struct http_txn *txn = s->txn; + struct htx *htx; + struct http_hdr_ctx ctx = { .blk = NULL }; + int pragma_found, cc_found; + + if ((txn->flags & (TX_CACHEABLE|TX_CACHE_IGNORE)) == TX_CACHE_IGNORE) + return; /* nothing more to do here */ + + htx = htxbuf(&req->buf); + pragma_found = cc_found = 0; + + /* Check "pragma" header for HTTP/1.0 compatibility. */ + if (http_find_header(htx, ist("pragma"), &ctx, 1)) { + if (isteqi(ctx.value, ist("no-cache"))) { + pragma_found = 1; + } + } + + ctx.blk = NULL; + /* Don't use the cache and don't try to store if we found the + * Authorization header */ + if (http_find_header(htx, ist("authorization"), &ctx, 1)) { + txn->flags &= ~TX_CACHEABLE & ~TX_CACHE_COOK; + txn->flags |= TX_CACHE_IGNORE; + } + + + /* Look for "cache-control" header and iterate over all the values + * until we find one that specifies that caching is possible or not. */ + ctx.blk = NULL; + while (http_find_header(htx, ist("cache-control"), &ctx, 0)) { + cc_found = 1; + /* We don't check the values after max-age, max-stale nor min-fresh, + * we simply don't use the cache when they're specified. */ + if (istmatchi(ctx.value, ist("max-age")) || + istmatchi(ctx.value, ist("no-cache")) || + istmatchi(ctx.value, ist("max-stale")) || + istmatchi(ctx.value, ist("min-fresh"))) { + txn->flags |= TX_CACHE_IGNORE; + continue; + } + if (istmatchi(ctx.value, ist("no-store"))) { + txn->flags &= ~TX_CACHEABLE & ~TX_CACHE_COOK; + continue; + } + } + + /* RFC7234#5.4: + * When the Cache-Control header field is also present and + * understood in a request, Pragma is ignored. + * When the Cache-Control header field is not present in a + * request, caches MUST consider the no-cache request + * pragma-directive as having the same effect as if + * "Cache-Control: no-cache" were present. + */ + if (!cc_found && pragma_found) + txn->flags |= TX_CACHE_IGNORE; +} + +/* + * Check if response is cacheable or not. Updates s->txn->flags. + */ +void http_check_response_for_cacheability(struct stream *s, struct channel *res) +{ + struct http_txn *txn = s->txn; + struct http_hdr_ctx ctx = { .blk = NULL }; + struct htx *htx; + int has_freshness_info = 0; + int has_validator = 0; + + if (txn->status < 200) { + /* do not try to cache interim responses! */ + txn->flags &= ~TX_CACHEABLE & ~TX_CACHE_COOK; + return; + } + + htx = htxbuf(&res->buf); + /* Check "pragma" header for HTTP/1.0 compatibility. */ + if (http_find_header(htx, ist("pragma"), &ctx, 1)) { + if (isteqi(ctx.value, ist("no-cache"))) { + txn->flags &= ~TX_CACHEABLE & ~TX_CACHE_COOK; + return; + } + } + + /* Look for "cache-control" header and iterate over all the values + * until we find one that specifies that caching is possible or not. */ + ctx.blk = NULL; + while (http_find_header(htx, ist("cache-control"), &ctx, 0)) { + if (isteqi(ctx.value, ist("public"))) { + txn->flags |= TX_CACHEABLE | TX_CACHE_COOK; + continue; + } + if (isteqi(ctx.value, ist("private")) || + isteqi(ctx.value, ist("no-cache")) || + isteqi(ctx.value, ist("no-store")) || + isteqi(ctx.value, ist("max-age=0")) || + isteqi(ctx.value, ist("s-maxage=0"))) { + txn->flags &= ~TX_CACHEABLE & ~TX_CACHE_COOK; + continue; + } + /* We might have a no-cache="set-cookie" form. */ + if (istmatchi(ctx.value, ist("no-cache=\"set-cookie"))) { + txn->flags &= ~TX_CACHE_COOK; + continue; + } + + if (istmatchi(ctx.value, ist("s-maxage")) || + istmatchi(ctx.value, ist("max-age"))) { + has_freshness_info = 1; + continue; + } + } + + /* If no freshness information could be found in Cache-Control values, + * look for an Expires header. */ + if (!has_freshness_info) { + ctx.blk = NULL; + has_freshness_info = http_find_header(htx, ist("expires"), &ctx, 0); + } + + /* If no freshness information could be found in Cache-Control or Expires + * values, look for an explicit validator. */ + if (!has_freshness_info) { + ctx.blk = NULL; + has_validator = 1; + if (!http_find_header(htx, ist("etag"), &ctx, 0)) { + ctx.blk = NULL; + if (!http_find_header(htx, ist("last-modified"), &ctx, 0)) + has_validator = 0; + } + } + + /* We won't store an entry that has neither a cache validator nor an + * explicit expiration time, as suggested in RFC 7234#3. */ + if (!has_freshness_info && !has_validator) + txn->flags &= ~TX_CACHEABLE; +} + +/* + * In a GET, HEAD or POST request, check if the requested URI matches the stats uri + * for the current backend. + * + * It is assumed that the request is either a HEAD, GET, or POST and that the + * uri_auth field is valid. + * + * Returns 1 if stats should be provided, otherwise 0. + */ +static int http_stats_check_uri(struct stream *s, struct http_txn *txn, struct proxy *backend) +{ + struct uri_auth *uri_auth = backend->uri_auth; + struct htx *htx; + struct htx_sl *sl; + struct ist uri; + + if (!uri_auth) + return 0; + + if (txn->meth != HTTP_METH_GET && txn->meth != HTTP_METH_HEAD && txn->meth != HTTP_METH_POST) + return 0; + + htx = htxbuf(&s->req.buf); + sl = http_get_stline(htx); + uri = htx_sl_req_uri(sl); + if (*uri_auth->uri_prefix == '/') { + struct http_uri_parser parser = http_uri_parser_init(uri); + uri = http_parse_path(&parser); + } + + /* check URI size */ + if (uri_auth->uri_len > uri.len) + return 0; + + if (memcmp(uri.ptr, uri_auth->uri_prefix, uri_auth->uri_len) != 0) + return 0; + + return 1; +} + +/* This function prepares an applet to handle the stats. It can deal with the + * "100-continue" expectation, check that admin rules are met for POST requests, + * and program a response message if something was unexpected. It cannot fail + * and always relies on the stats applet to complete the job. It does not touch + * analysers nor counters, which are left to the caller. It does not touch + * s->target which is supposed to already point to the stats applet. The caller + * is expected to have already assigned an appctx to the stream. + */ +static int http_handle_stats(struct stream *s, struct channel *req) +{ + struct stats_admin_rule *stats_admin_rule; + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct http_msg *msg = &txn->req; + struct uri_auth *uri_auth = s->be->uri_auth; + const char *h, *lookup, *end; + struct appctx *appctx = __sc_appctx(s->scb); + struct show_stat_ctx *ctx = applet_reserve_svcctx(appctx, sizeof(*ctx)); + struct htx *htx; + struct htx_sl *sl; + + appctx->st1 = 0; + ctx->state = STAT_STATE_INIT; + ctx->st_code = STAT_STATUS_INIT; + ctx->flags |= uri_auth->flags; + ctx->flags |= STAT_FMT_HTML; /* assume HTML mode by default */ + if ((msg->flags & HTTP_MSGF_VER_11) && (txn->meth != HTTP_METH_HEAD)) + ctx->flags |= STAT_CHUNKED; + + htx = htxbuf(&req->buf); + sl = http_get_stline(htx); + lookup = HTX_SL_REQ_UPTR(sl) + uri_auth->uri_len; + end = HTX_SL_REQ_UPTR(sl) + HTX_SL_REQ_ULEN(sl); + + for (h = lookup; h <= end - 3; h++) { + if (memcmp(h, ";up", 3) == 0) { + ctx->flags |= STAT_HIDE_DOWN; + break; + } + } + + for (h = lookup; h <= end - 9; h++) { + if (memcmp(h, ";no-maint", 9) == 0) { + ctx->flags |= STAT_HIDE_MAINT; + break; + } + } + + if (uri_auth->refresh) { + for (h = lookup; h <= end - 10; h++) { + if (memcmp(h, ";norefresh", 10) == 0) { + ctx->flags |= STAT_NO_REFRESH; + break; + } + } + } + + for (h = lookup; h <= end - 4; h++) { + if (memcmp(h, ";csv", 4) == 0) { + ctx->flags &= ~(STAT_FMT_MASK|STAT_JSON_SCHM); + break; + } + } + + for (h = lookup; h <= end - 6; h++) { + if (memcmp(h, ";typed", 6) == 0) { + ctx->flags &= ~(STAT_FMT_MASK|STAT_JSON_SCHM); + ctx->flags |= STAT_FMT_TYPED; + break; + } + } + + for (h = lookup; h <= end - 5; h++) { + if (memcmp(h, ";json", 5) == 0) { + ctx->flags &= ~(STAT_FMT_MASK|STAT_JSON_SCHM); + ctx->flags |= STAT_FMT_JSON; + break; + } + } + + for (h = lookup; h <= end - 12; h++) { + if (memcmp(h, ";json-schema", 12) == 0) { + ctx->flags &= ~STAT_FMT_MASK; + ctx->flags |= STAT_JSON_SCHM; + break; + } + } + + for (h = lookup; h <= end - 8; h++) { + if (memcmp(h, ";st=", 4) == 0) { + int i; + h += 4; + ctx->st_code = STAT_STATUS_UNKN; + for (i = STAT_STATUS_INIT + 1; i < STAT_STATUS_SIZE; i++) { + if (strncmp(stat_status_codes[i], h, 4) == 0) { + ctx->st_code = i; + break; + } + } + break; + } + } + + ctx->scope_str = 0; + ctx->scope_len = 0; + for (h = lookup; h <= end - 8; h++) { + if (memcmp(h, STAT_SCOPE_INPUT_NAME "=", strlen(STAT_SCOPE_INPUT_NAME) + 1) == 0) { + int itx = 0; + const char *h2; + char scope_txt[STAT_SCOPE_TXT_MAXLEN + 1]; + const char *err; + + h += strlen(STAT_SCOPE_INPUT_NAME) + 1; + h2 = h; + ctx->scope_str = h2 - HTX_SL_REQ_UPTR(sl); + while (h < end) { + if (*h == ';' || *h == '&' || *h == ' ') + break; + itx++; + h++; + } + + if (itx > STAT_SCOPE_TXT_MAXLEN) + itx = STAT_SCOPE_TXT_MAXLEN; + ctx->scope_len = itx; + + /* scope_txt = search query, ctx->scope_len is always <= STAT_SCOPE_TXT_MAXLEN */ + memcpy(scope_txt, h2, itx); + scope_txt[itx] = '\0'; + err = invalid_char(scope_txt); + if (err) { + /* bad char in search text => clear scope */ + ctx->scope_str = 0; + ctx->scope_len = 0; + } + break; + } + } + + /* now check whether we have some admin rules for this request */ + list_for_each_entry(stats_admin_rule, &uri_auth->admin_rules, list) { + int ret = 1; + + if (stats_admin_rule->cond) { + ret = acl_exec_cond(stats_admin_rule->cond, s->be, sess, s, SMP_OPT_DIR_REQ|SMP_OPT_FINAL); + ret = acl_pass(ret); + if (stats_admin_rule->cond->pol == ACL_COND_UNLESS) + ret = !ret; + } + + if (ret) { + /* no rule, or the rule matches */ + ctx->flags |= STAT_ADMIN; + break; + } + } + + if (txn->meth == HTTP_METH_GET || txn->meth == HTTP_METH_HEAD) + appctx->st0 = STAT_HTTP_HEAD; + else if (txn->meth == HTTP_METH_POST) { + if (ctx->flags & STAT_ADMIN) { + appctx->st0 = STAT_HTTP_POST; + if (msg->msg_state < HTTP_MSG_DATA) + req->analysers |= AN_REQ_HTTP_BODY; + } + else { + /* POST without admin level */ + ctx->flags &= ~STAT_CHUNKED; + ctx->st_code = STAT_STATUS_DENY; + appctx->st0 = STAT_HTTP_LAST; + } + } + else { + /* Unsupported method */ + ctx->flags &= ~STAT_CHUNKED; + ctx->st_code = STAT_STATUS_IVAL; + appctx->st0 = STAT_HTTP_LAST; + } + + s->task->nice = -32; /* small boost for HTTP statistics */ + return 1; +} + +/* This function waits for the message payload at most <time> milliseconds (may + * be set to TICK_ETERNITY). It stops to wait if at least <bytes> bytes of the + * payload are received (0 means no limit). It returns HTTP_RULE_* depending on + * the result: + * + * - HTTP_RULE_RES_CONT when conditions are met to stop waiting + * - HTTP_RULE_RES_YIELD to wait for more data + * - HTTP_RULE_RES_ABRT when a timeout occurred. + * - HTTP_RULE_RES_BADREQ if a parsing error is raised by lower level + * - HTTP_RULE_RES_ERROR if an internal error occurred + * + * If a timeout occurred, this function is responsible to emit the right response + * to the client, depending on the channel (408 on request side, 504 on response + * side). All other errors must be handled by the caller. + */ +enum rule_result http_wait_for_msg_body(struct stream *s, struct channel *chn, + unsigned int time, unsigned int bytes) +{ + struct session *sess = s->sess; + struct http_txn *txn = s->txn; + struct http_msg *msg = ((chn->flags & CF_ISRESP) ? &txn->rsp : &txn->req); + struct htx *htx; + enum rule_result ret = HTTP_RULE_RES_CONT; + + htx = htxbuf(&chn->buf); + + if (htx->flags & HTX_FL_PARSING_ERROR) { + ret = HTTP_RULE_RES_BADREQ; + goto end; + } + if (htx->flags & HTX_FL_PROCESSING_ERROR) { + ret = HTTP_RULE_RES_ERROR; + goto end; + } + + /* Do nothing for bodyless and CONNECT requests */ + if (txn->meth == HTTP_METH_CONNECT || (msg->flags & HTTP_MSGF_BODYLESS)) + goto end; + + if (!(chn->flags & CF_ISRESP) && msg->msg_state < HTTP_MSG_DATA) { + if (http_handle_expect_hdr(s, htx, msg) == -1) { + ret = HTTP_RULE_RES_ERROR; + goto end; + } + } + + msg->msg_state = HTTP_MSG_DATA; + + /* Now we're in HTTP_MSG_DATA. We just need to know if all data have + * been received or if the buffer is full. + */ + if ((htx->flags & HTX_FL_EOM) || + htx_get_tail_type(htx) > HTX_BLK_DATA || + channel_htx_full(chn, htx, global.tune.maxrewrite) || + sc_waiting_room(chn_prod(chn))) + goto end; + + if (bytes) { + struct htx_blk *blk; + unsigned int len = 0; + + for (blk = htx_get_first_blk(htx); blk; blk = htx_get_next_blk(htx, blk)) { + if (htx_get_blk_type(blk) != HTX_BLK_DATA) + continue; + len += htx_get_blksz(blk); + if (len >= bytes) + goto end; + } + } + + if ((chn->flags & CF_READ_TIMEOUT) || tick_is_expired(chn->analyse_exp, now_ms)) { + if (!(chn->flags & CF_ISRESP)) + goto abort_req; + goto abort_res; + } + + /* we get here if we need to wait for more data */ + if (!(chn->flags & (CF_SHUTR | CF_READ_ERROR))) { + if (!tick_isset(chn->analyse_exp)) + chn->analyse_exp = tick_add_ifset(now_ms, time); + ret = HTTP_RULE_RES_YIELD; + } + + end: + return ret; + + abort_req: + txn->status = 408; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_CLITO; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_R; + _HA_ATOMIC_INC(&sess->fe->fe_counters.failed_req); + if (sess->listener && sess->listener->counters) + _HA_ATOMIC_INC(&sess->listener->counters->failed_req); + http_reply_and_close(s, txn->status, http_error_message(s)); + ret = HTTP_RULE_RES_ABRT; + goto end; + + abort_res: + txn->status = 504; + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_SRVTO; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_R; + stream_inc_http_fail_ctr(s); + http_reply_and_close(s, txn->status, http_error_message(s)); + ret = HTTP_RULE_RES_ABRT; + goto end; +} + +void http_perform_server_redirect(struct stream *s, struct stconn *sc) +{ + struct channel *req = &s->req; + struct channel *res = &s->res; + struct server *srv; + struct htx *htx; + struct htx_sl *sl; + struct ist path, location; + unsigned int flags; + struct http_uri_parser parser; + + /* + * Create the location + */ + chunk_reset(&trash); + + /* 1: add the server's prefix */ + /* special prefix "/" means don't change URL */ + srv = __objt_server(s->target); + if (srv->rdr_len != 1 || *srv->rdr_pfx != '/') { + if (!chunk_memcat(&trash, srv->rdr_pfx, srv->rdr_len)) + return; + } + + /* 2: add the request Path */ + htx = htxbuf(&req->buf); + sl = http_get_stline(htx); + parser = http_uri_parser_init(htx_sl_req_uri(sl)); + path = http_parse_path(&parser); + if (!isttest(path)) + return; + + if (!chunk_memcat(&trash, path.ptr, path.len)) + return; + location = ist2(trash.area, trash.data); + + /* + * Create the 302 respone + */ + htx = htx_from_buf(&res->buf); + flags = (HTX_SL_F_IS_RESP|HTX_SL_F_VER_11|HTX_SL_F_XFER_LEN|HTX_SL_F_BODYLESS); + sl = htx_add_stline(htx, HTX_BLK_RES_SL, flags, + ist("HTTP/1.1"), ist("302"), ist("Found")); + if (!sl) + goto fail; + sl->info.res.status = 302; + s->txn->status = 302; + + if (!htx_add_header(htx, ist("Cache-Control"), ist("no-cache")) || + !htx_add_header(htx, ist("Connection"), ist("close")) || + !htx_add_header(htx, ist("Content-length"), ist("0")) || + !htx_add_header(htx, ist("Location"), location)) + goto fail; + + if (!htx_add_endof(htx, HTX_BLK_EOH)) + goto fail; + + htx->flags |= HTX_FL_EOM; + htx_to_buf(htx, &res->buf); + if (!http_forward_proxy_resp(s, 1)) + goto fail; + + /* return without error. */ + sc_shutr(sc); + sc_shutw(sc); + s->conn_err_type = STRM_ET_NONE; + sc->state = SC_ST_CLO; + + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_LOCAL; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= SF_FINST_C; + + /* FIXME: we should increase a counter of redirects per server and per backend. */ + srv_inc_sess_ctr(srv); + srv_set_sess_last(srv); + return; + + fail: + /* If an error occurred, remove the incomplete HTTP response from the + * buffer */ + channel_htx_truncate(res, htx); +} + +/* This function terminates the request because it was completely analyzed or + * because an error was triggered during the body forwarding. + */ +static void http_end_request(struct stream *s) +{ + struct channel *chn = &s->req; + struct http_txn *txn = s->txn; + + DBG_TRACE_ENTER(STRM_EV_HTTP_ANA, s, txn); + + if (unlikely(txn->req.msg_state == HTTP_MSG_ERROR || + txn->rsp.msg_state == HTTP_MSG_ERROR)) { + channel_abort(chn); + channel_htx_truncate(chn, htxbuf(&chn->buf)); + goto end; + } + + if (unlikely(txn->req.msg_state < HTTP_MSG_DONE)) { + DBG_TRACE_DEVEL("waiting end of the request", STRM_EV_HTTP_ANA, s, txn); + return; + } + + if (txn->req.msg_state == HTTP_MSG_DONE) { + /* No need to read anymore, the request was completely parsed. + * We can shut the read side unless we want to abort_on_close, + * or we have a POST request. The issue with POST requests is + * that some browsers still send a CRLF after the request, and + * this CRLF must be read so that it does not remain in the kernel + * buffers, otherwise a close could cause an RST on some systems + * (eg: Linux). + */ + if (!(s->be->options & PR_O_ABRT_CLOSE) && txn->meth != HTTP_METH_POST) + channel_dont_read(chn); + + /* if the server closes the connection, we want to immediately react + * and close the socket to save packets and syscalls. + */ + s->scb->flags |= SC_FL_NOHALF; + + /* In any case we've finished parsing the request so we must + * disable Nagle when sending data because 1) we're not going + * to shut this side, and 2) the server is waiting for us to + * send pending data. + */ + chn->flags |= CF_NEVER_WAIT; + + if (txn->rsp.msg_state < HTTP_MSG_DONE) { + /* The server has not finished to respond, so we + * don't want to move in order not to upset it. + */ + DBG_TRACE_DEVEL("waiting end of the response", STRM_EV_HTTP_ANA, s, txn); + return; + } + + /* When we get here, it means that both the request and the + * response have finished receiving. Depending on the connection + * mode, we'll have to wait for the last bytes to leave in either + * direction, and sometimes for a close to be effective. + */ + if (txn->flags & TX_CON_WANT_TUN) { + /* Tunnel mode will not have any analyser so it needs to + * poll for reads. + */ + channel_auto_read(chn); + txn->req.msg_state = HTTP_MSG_TUNNEL; + } + else { + /* we're not expecting any new data to come for this + * transaction, so we can close it. + * + * However, there is an exception if the response + * length is undefined. In this case, we need to wait + * the close from the server. The response will be + * switched in TUNNEL mode until the end. + */ + if (!(txn->rsp.flags & HTTP_MSGF_XFER_LEN) && + txn->rsp.msg_state != HTTP_MSG_CLOSED) + goto check_channel_flags; + + if (!(chn->flags & (CF_SHUTW|CF_SHUTW_NOW))) { + channel_shutr_now(chn); + channel_shutw_now(chn); + } + } + goto check_channel_flags; + } + + if (txn->req.msg_state == HTTP_MSG_CLOSING) { + http_msg_closing: + /* nothing else to forward, just waiting for the output buffer + * to be empty and for the shutw_now to take effect. + */ + if (channel_is_empty(chn)) { + txn->req.msg_state = HTTP_MSG_CLOSED; + goto http_msg_closed; + } + else if (chn->flags & CF_SHUTW) { + txn->req.msg_state = HTTP_MSG_ERROR; + goto end; + } + DBG_TRACE_LEAVE(STRM_EV_HTTP_ANA, s, txn); + return; + } + + if (txn->req.msg_state == HTTP_MSG_CLOSED) { + http_msg_closed: + /* if we don't know whether the server will close, we need to hard close */ + if (txn->rsp.flags & HTTP_MSGF_XFER_LEN) + s->scb->flags |= SC_FL_NOLINGER; /* we want to close ASAP */ + /* see above in MSG_DONE why we only do this in these states */ + if (!(s->be->options & PR_O_ABRT_CLOSE)) + channel_dont_read(chn); + goto end; + } + + check_channel_flags: + /* Here, we are in HTTP_MSG_DONE or HTTP_MSG_TUNNEL */ + if (chn->flags & (CF_SHUTW|CF_SHUTW_NOW)) { + /* if we've just closed an output, let's switch */ + txn->req.msg_state = HTTP_MSG_CLOSING; + goto http_msg_closing; + } + + end: + chn->analysers &= AN_REQ_FLT_END; + if (txn->req.msg_state == HTTP_MSG_TUNNEL) { + chn->flags |= CF_NEVER_WAIT; + if (HAS_REQ_DATA_FILTERS(s)) + chn->analysers |= AN_REQ_FLT_XFER_DATA; + } + channel_auto_close(chn); + channel_auto_read(chn); + DBG_TRACE_LEAVE(STRM_EV_HTTP_ANA, s, txn); +} + + +/* This function terminates the response because it was completely analyzed or + * because an error was triggered during the body forwarding. + */ +static void http_end_response(struct stream *s) +{ + struct channel *chn = &s->res; + struct http_txn *txn = s->txn; + + DBG_TRACE_ENTER(STRM_EV_HTTP_ANA, s, txn); + + if (unlikely(txn->req.msg_state == HTTP_MSG_ERROR || + txn->rsp.msg_state == HTTP_MSG_ERROR)) { + channel_htx_truncate(&s->req, htxbuf(&s->req.buf)); + channel_abort(&s->req); + goto end; + } + + if (unlikely(txn->rsp.msg_state < HTTP_MSG_DONE)) { + DBG_TRACE_DEVEL("waiting end of the response", STRM_EV_HTTP_ANA, s, txn); + return; + } + + if (txn->rsp.msg_state == HTTP_MSG_DONE) { + /* In theory, we don't need to read anymore, but we must + * still monitor the server connection for a possible close + * while the request is being uploaded, so we don't disable + * reading. + */ + /* channel_dont_read(chn); */ + + if (txn->req.msg_state < HTTP_MSG_DONE) { + /* The client seems to still be sending data, probably + * because we got an error response during an upload. + * We have the choice of either breaking the connection + * or letting it pass through. Let's do the later. + */ + DBG_TRACE_DEVEL("waiting end of the request", STRM_EV_HTTP_ANA, s, txn); + return; + } + + /* When we get here, it means that both the request and the + * response have finished receiving. Depending on the connection + * mode, we'll have to wait for the last bytes to leave in either + * direction, and sometimes for a close to be effective. + */ + if (txn->flags & TX_CON_WANT_TUN) { + channel_auto_read(chn); + txn->rsp.msg_state = HTTP_MSG_TUNNEL; + } + else { + /* we're not expecting any new data to come for this + * transaction, so we can close it. + */ + if (!(chn->flags & (CF_SHUTW|CF_SHUTW_NOW))) { + channel_shutr_now(chn); + channel_shutw_now(chn); + } + } + goto check_channel_flags; + } + + if (txn->rsp.msg_state == HTTP_MSG_CLOSING) { + http_msg_closing: + /* nothing else to forward, just waiting for the output buffer + * to be empty and for the shutw_now to take effect. + */ + if (channel_is_empty(chn)) { + txn->rsp.msg_state = HTTP_MSG_CLOSED; + goto http_msg_closed; + } + else if (chn->flags & CF_SHUTW) { + txn->rsp.msg_state = HTTP_MSG_ERROR; + _HA_ATOMIC_INC(&strm_sess(s)->fe->fe_counters.cli_aborts); + _HA_ATOMIC_INC(&s->be->be_counters.cli_aborts); + if (strm_sess(s)->listener && strm_sess(s)->listener->counters) + _HA_ATOMIC_INC(&strm_sess(s)->listener->counters->cli_aborts); + if (objt_server(s->target)) + _HA_ATOMIC_INC(&__objt_server(s->target)->counters.cli_aborts); + goto end; + } + DBG_TRACE_LEAVE(STRM_EV_HTTP_ANA, s, txn); + return; + } + + if (txn->rsp.msg_state == HTTP_MSG_CLOSED) { + http_msg_closed: + /* drop any pending data */ + channel_htx_truncate(&s->req, htxbuf(&s->req.buf)); + channel_abort(&s->req); + goto end; + } + + check_channel_flags: + /* Here, we are in HTTP_MSG_DONE or HTTP_MSG_TUNNEL */ + if (chn->flags & (CF_SHUTW|CF_SHUTW_NOW)) { + /* if we've just closed an output, let's switch */ + txn->rsp.msg_state = HTTP_MSG_CLOSING; + goto http_msg_closing; + } + + end: + chn->analysers &= AN_RES_FLT_END; + if (txn->rsp.msg_state == HTTP_MSG_TUNNEL) { + chn->flags |= CF_NEVER_WAIT; + if (HAS_RSP_DATA_FILTERS(s)) + chn->analysers |= AN_RES_FLT_XFER_DATA; + } + channel_auto_close(chn); + channel_auto_read(chn); + DBG_TRACE_LEAVE(STRM_EV_HTTP_ANA, s, txn); +} + +/* Forward a response generated by HAProxy (error/redirect/return). This + * function forwards all pending incoming data. If <final> is set to 0, nothing + * more is performed. It is used for 1xx informational messages. Otherwise, the + * transaction is terminated and the request is emptied. On success 1 is + * returned. If an error occurred, 0 is returned. If it fails, this function + * only exits. It is the caller responsibility to do the cleanup. + */ +int http_forward_proxy_resp(struct stream *s, int final) +{ + struct channel *req = &s->req; + struct channel *res = &s->res; + struct htx *htx = htxbuf(&res->buf); + size_t data; + + if (final) { + htx->flags |= HTX_FL_PROXY_RESP; + + if (!htx_is_empty(htx) && !http_eval_after_res_rules(s)) + return 0; + + if (s->txn->meth == HTTP_METH_HEAD) + htx_skip_msg_payload(htx); + + channel_auto_read(req); + channel_abort(req); + channel_auto_close(req); + channel_htx_erase(req, htxbuf(&req->buf)); + + res->wex = tick_add_ifset(now_ms, res->wto); + channel_auto_read(res); + channel_auto_close(res); + channel_shutr_now(res); + res->flags |= CF_EOI; /* The response is terminated, add EOI */ + htxbuf(&res->buf)->flags |= HTX_FL_EOM; /* no more data are expected */ + } + else { + /* Send ASAP informational messages. Rely on CF_EOI for final + * response. + */ + res->flags |= CF_SEND_DONTWAIT; + } + + data = htx->data - co_data(res); + c_adv(res, data); + htx->first = -1; + res->total += data; + return 1; +} + +void http_server_error(struct stream *s, struct stconn *sc, int err, + int finst, struct http_reply *msg) +{ + http_reply_and_close(s, s->txn->status, msg); + if (!(s->flags & SF_ERR_MASK)) + s->flags |= err; + if (!(s->flags & SF_FINST_MASK)) + s->flags |= finst; +} + +void http_reply_and_close(struct stream *s, short status, struct http_reply *msg) +{ + if (!msg) { + channel_htx_truncate(&s->res, htxbuf(&s->res.buf)); + goto end; + } + + if (http_reply_message(s, msg) == -1) { + /* On error, return a 500 error message, but don't rewrite it if + * it is already an internal error. If it was already a "const" + * 500 error, just fail. + */ + if (s->txn->status == 500) { + if (s->txn->flags & TX_CONST_REPLY) + goto end; + s->txn->flags |= TX_CONST_REPLY; + } + s->txn->status = 500; + s->txn->http_reply = NULL; + return http_reply_and_close(s, s->txn->status, http_error_message(s)); + } + +end: + s->res.wex = tick_add_ifset(now_ms, s->res.wto); + + /* At this staged, HTTP analysis is finished */ + s->req.analysers &= AN_REQ_FLT_END; + s->req.analyse_exp = TICK_ETERNITY; + + s->res.analysers &= AN_RES_FLT_END; + s->res.analyse_exp = TICK_ETERNITY; + + channel_auto_read(&s->req); + channel_abort(&s->req); + channel_auto_close(&s->req); + channel_htx_erase(&s->req, htxbuf(&s->req.buf)); + channel_auto_read(&s->res); + channel_auto_close(&s->res); + channel_shutr_now(&s->res); +} + +struct http_reply *http_error_message(struct stream *s) +{ + const int msgnum = http_get_status_idx(s->txn->status); + + if (s->txn->http_reply) + return s->txn->http_reply; + else if (s->be->replies[msgnum]) + return s->be->replies[msgnum]; + else if (strm_fe(s)->replies[msgnum]) + return strm_fe(s)->replies[msgnum]; + else + return &http_err_replies[msgnum]; +} + +/* Produces an HTX message from an http reply. Depending on the http reply type, + * a, errorfile, an raw file or a log-format string is used. On success, it + * returns 0. If an error occurs -1 is returned. If it fails, this function only + * exits. It is the caller responsibility to do the cleanup. + */ +int http_reply_to_htx(struct stream *s, struct htx *htx, struct http_reply *reply) +{ + struct buffer *errmsg; + struct htx_sl *sl; + struct buffer *body = NULL; + const char *status, *reason, *clen, *ctype; + unsigned int slflags; + int ret = 0; + + /* + * - HTTP_REPLY_ERRFILES unexpected here. handled as no payload if so + * + * - HTTP_REPLY_INDIRECT: switch on another reply if defined or handled + * as no payload if NULL. the TXN status code is set with the status + * of the original reply. + */ + + if (reply->type == HTTP_REPLY_INDIRECT) { + if (reply->body.reply) + reply = reply->body.reply; + } + if (reply->type == HTTP_REPLY_ERRMSG && !reply->body.errmsg) { + /* get default error message */ + if (reply == s->txn->http_reply) + s->txn->http_reply = NULL; + reply = http_error_message(s); + if (reply->type == HTTP_REPLY_INDIRECT) { + if (reply->body.reply) + reply = reply->body.reply; + } + } + + if (reply->type == HTTP_REPLY_ERRMSG) { + /* implicit or explicit error message*/ + errmsg = reply->body.errmsg; + if (errmsg && !b_is_null(errmsg)) { + if (!htx_copy_msg(htx, errmsg)) + goto fail; + } + } + else { + /* no payload, file or log-format string */ + if (reply->type == HTTP_REPLY_RAW) { + /* file */ + body = &reply->body.obj; + } + else if (reply->type == HTTP_REPLY_LOGFMT) { + /* log-format string */ + body = alloc_trash_chunk(); + if (!body) + goto fail_alloc; + body->data = build_logline(s, body->area, body->size, &reply->body.fmt); + } + /* else no payload */ + + status = ultoa(reply->status); + reason = http_get_reason(reply->status); + slflags = (HTX_SL_F_IS_RESP|HTX_SL_F_VER_11|HTX_SL_F_XFER_LEN|HTX_SL_F_CLEN); + if (!body || !b_data(body)) + slflags |= HTX_SL_F_BODYLESS; + sl = htx_add_stline(htx, HTX_BLK_RES_SL, slflags, ist("HTTP/1.1"), ist(status), ist(reason)); + if (!sl) + goto fail; + sl->info.res.status = reply->status; + + clen = (body ? ultoa(b_data(body)) : "0"); + ctype = reply->ctype; + + if (!LIST_ISEMPTY(&reply->hdrs)) { + struct http_reply_hdr *hdr; + struct buffer *value = alloc_trash_chunk(); + + if (!value) + goto fail; + + list_for_each_entry(hdr, &reply->hdrs, list) { + chunk_reset(value); + value->data = build_logline(s, value->area, value->size, &hdr->value); + if (b_data(value) && !htx_add_header(htx, hdr->name, ist2(b_head(value), b_data(value)))) { + free_trash_chunk(value); + goto fail; + } + chunk_reset(value); + } + free_trash_chunk(value); + } + + if (!htx_add_header(htx, ist("content-length"), ist(clen)) || + (body && b_data(body) && ctype && !htx_add_header(htx, ist("content-type"), ist(ctype))) || + !htx_add_endof(htx, HTX_BLK_EOH) || + (body && b_data(body) && !htx_add_data_atonce(htx, ist2(b_head(body), b_data(body))))) + goto fail; + + htx->flags |= HTX_FL_EOM; + } + + leave: + if (reply->type == HTTP_REPLY_LOGFMT) + free_trash_chunk(body); + return ret; + + fail_alloc: + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_RESOURCE; + /* fall through */ + fail: + ret = -1; + goto leave; +} + +/* Send an http reply to the client. On success, it returns 0. If an error + * occurs -1 is returned and the response channel is truncated, removing this + * way the faulty reply. This function may fail when the reply is formatted + * (http_reply_to_htx) or when the reply is forwarded + * (http_forward_proxy_resp). On the last case, it is because a + * http-after-response rule fails. + */ +int http_reply_message(struct stream *s, struct http_reply *reply) +{ + struct channel *res = &s->res; + struct htx *htx = htx_from_buf(&res->buf); + + if (s->txn->status == -1) + s->txn->status = reply->status; + channel_htx_truncate(res, htx); + + if (http_reply_to_htx(s, htx, reply) == -1) + goto fail; + + htx_to_buf(htx, &s->res.buf); + if (!http_forward_proxy_resp(s, 1)) + goto fail; + return 0; + + fail: + channel_htx_truncate(res, htx); + if (!(s->flags & SF_ERR_MASK)) + s->flags |= SF_ERR_PRXCOND; + return -1; +} + +/* Return the error message corresponding to s->conn_err_type. It is assumed + * that the server side is closed. Note that err_type is actually a + * bitmask, where almost only aborts may be cumulated with other + * values. We consider that aborted operations are more important + * than timeouts or errors due to the fact that nobody else in the + * logs might explain incomplete retries. All others should avoid + * being cumulated. It should normally not be possible to have multiple + * aborts at once, but just in case, the first one in sequence is reported. + * Note that connection errors appearing on the second request of a keep-alive + * connection are not reported since this allows the client to retry. + */ +void http_return_srv_error(struct stream *s, struct stconn *sc) +{ + int err_type = s->conn_err_type; + + /* set s->txn->status for http_error_message(s) */ + if (err_type & STRM_ET_QUEUE_ABRT) { + s->txn->status = -1; + http_server_error(s, sc, SF_ERR_CLICL, SF_FINST_Q, NULL); + } + else if (err_type & STRM_ET_CONN_ABRT) { + s->txn->status = -1; + http_server_error(s, sc, SF_ERR_CLICL, SF_FINST_C, NULL); + } + else if (err_type & STRM_ET_QUEUE_TO) { + s->txn->status = 503; + http_server_error(s, sc, SF_ERR_SRVTO, SF_FINST_Q, + http_error_message(s)); + } + else if (err_type & STRM_ET_QUEUE_ERR) { + s->txn->status = 503; + http_server_error(s, sc, SF_ERR_SRVCL, SF_FINST_Q, + http_error_message(s)); + } + else if (err_type & STRM_ET_CONN_TO) { + s->txn->status = 503; + http_server_error(s, sc, SF_ERR_SRVTO, SF_FINST_C, + (s->txn->flags & TX_NOT_FIRST) ? NULL : + http_error_message(s)); + } + else if (err_type & STRM_ET_CONN_ERR) { + s->txn->status = 503; + http_server_error(s, sc, SF_ERR_SRVCL, SF_FINST_C, + (s->flags & SF_SRV_REUSED) ? NULL : + http_error_message(s)); + } + else if (err_type & STRM_ET_CONN_RES) { + s->txn->status = 503; + http_server_error(s, sc, SF_ERR_RESOURCE, SF_FINST_C, + (s->txn->flags & TX_NOT_FIRST) ? NULL : + http_error_message(s)); + } + else { /* STRM_ET_CONN_OTHER and others */ + s->txn->status = 500; + http_server_error(s, sc, SF_ERR_INTERNAL, SF_FINST_C, + http_error_message(s)); + } +} + + +/* Handle Expect: 100-continue for HTTP/1.1 messages if necessary. It returns 0 + * on success and -1 on error. + */ +static int http_handle_expect_hdr(struct stream *s, struct htx *htx, struct http_msg *msg) +{ + /* If we have HTTP/1.1 message with a body and Expect: 100-continue, + * then we must send an HTTP/1.1 100 Continue intermediate response. + */ + if (msg->msg_state == HTTP_MSG_BODY && (msg->flags & HTTP_MSGF_VER_11) && + (msg->flags & (HTTP_MSGF_CNT_LEN|HTTP_MSGF_TE_CHNK))) { + struct ist hdr = { .ptr = "Expect", .len = 6 }; + struct http_hdr_ctx ctx; + + ctx.blk = NULL; + /* Expect is allowed in 1.1, look for it */ + if (http_find_header(htx, hdr, &ctx, 0) && + unlikely(isteqi(ctx.value, ist2("100-continue", 12)))) { + if (http_reply_100_continue(s) == -1) + return -1; + http_remove_header(htx, &ctx); + } + } + return 0; +} + +/* Send a 100-Continue response to the client. It returns 0 on success and -1 + * on error. The response channel is updated accordingly. + */ +static int http_reply_100_continue(struct stream *s) +{ + struct channel *res = &s->res; + struct htx *htx = htx_from_buf(&res->buf); + struct htx_sl *sl; + unsigned int flags = (HTX_SL_F_IS_RESP|HTX_SL_F_VER_11| + HTX_SL_F_XFER_LEN|HTX_SL_F_BODYLESS); + + sl = htx_add_stline(htx, HTX_BLK_RES_SL, flags, + ist("HTTP/1.1"), ist("100"), ist("Continue")); + if (!sl) + goto fail; + sl->info.res.status = 100; + + if (!htx_add_endof(htx, HTX_BLK_EOH)) + goto fail; + + if (!http_forward_proxy_resp(s, 0)) + goto fail; + return 0; + + fail: + /* If an error occurred, remove the incomplete HTTP response from the + * buffer */ + channel_htx_truncate(res, htx); + return -1; +} + + +/* + * Capture headers from message <htx> according to header list <cap_hdr>, and + * fill the <cap> pointers appropriately. + */ +static void http_capture_headers(struct htx *htx, char **cap, struct cap_hdr *cap_hdr) +{ + struct cap_hdr *h; + int32_t pos; + + for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) { + struct htx_blk *blk = htx_get_blk(htx, pos); + enum htx_blk_type type = htx_get_blk_type(blk); + struct ist n, v; + + if (type == HTX_BLK_EOH) + break; + if (type != HTX_BLK_HDR) + continue; + + n = htx_get_blk_name(htx, blk); + + for (h = cap_hdr; h; h = h->next) { + if (h->namelen && (h->namelen == n.len) && + (strncasecmp(n.ptr, h->name, h->namelen) == 0)) { + if (cap[h->index] == NULL) + cap[h->index] = + pool_alloc(h->pool); + + if (cap[h->index] == NULL) { + ha_alert("HTTP capture : out of memory.\n"); + break; + } + + v = htx_get_blk_value(htx, blk); + v = isttrim(v, h->len); + + memcpy(cap[h->index], v.ptr, v.len); + cap[h->index][v.len]=0; + } + } + } +} + +/* Delete a value in a header between delimiters <from> and <next>. The header + * itself is delimited by <start> and <end> pointers. The number of characters + * displaced is returned, and the pointer to the first delimiter is updated if + * required. The function tries as much as possible to respect the following + * principles : + * - replace <from> delimiter by the <next> one unless <from> points to <start>, + * in which case <next> is simply removed + * - set exactly one space character after the new first delimiter, unless there + * are not enough characters in the block being moved to do so. + * - remove unneeded spaces before the previous delimiter and after the new + * one. + * + * It is the caller's responsibility to ensure that : + * - <from> points to a valid delimiter or <start> ; + * - <next> points to a valid delimiter or <end> ; + * - there are non-space chars before <from>. + */ +static int http_del_hdr_value(char *start, char *end, char **from, char *next) +{ + char *prev = *from; + + if (prev == start) { + /* We're removing the first value. eat the semicolon, if <next> + * is lower than <end> */ + if (next < end) + next++; + + while (next < end && HTTP_IS_SPHT(*next)) + next++; + } + else { + /* Remove useless spaces before the old delimiter. */ + while (HTTP_IS_SPHT(*(prev-1))) + prev--; + *from = prev; + + /* copy the delimiter and if possible a space if we're + * not at the end of the line. + */ + if (next < end) { + *prev++ = *next++; + if (prev + 1 < next) + *prev++ = ' '; + while (next < end && HTTP_IS_SPHT(*next)) + next++; + } + } + memmove(prev, next, end - next); + return (prev - next); +} + + +/* Formats the start line of the request (without CRLF) and puts it in <str> and + * return the written length. The line can be truncated if it exceeds <len>. + */ +static size_t http_fmt_req_line(const struct htx_sl *sl, char *str, size_t len) +{ + struct ist dst = ist2(str, 0); + + if (istcat(&dst, htx_sl_req_meth(sl), len) == -1) + goto end; + if (dst.len + 1 > len) + goto end; + dst.ptr[dst.len++] = ' '; + + if (istcat(&dst, htx_sl_req_uri(sl), len) == -1) + goto end; + if (dst.len + 1 > len) + goto end; + dst.ptr[dst.len++] = ' '; + + istcat(&dst, htx_sl_req_vsn(sl), len); + end: + return dst.len; +} + +/* + * Print a debug line with a start line. + */ +static void http_debug_stline(const char *dir, struct stream *s, const struct htx_sl *sl) +{ + struct session *sess = strm_sess(s); + int max; + + chunk_printf(&trash, "%08x:%s.%s[%04x:%04x]: ", s->uniq_id, s->be->id, + dir, + objt_conn(sess->origin) ? (unsigned short)__objt_conn(sess->origin)->handle.fd : -1, + sc_conn(s->scb) ? (unsigned short)(__sc_conn(s->scb))->handle.fd : -1); + + max = HTX_SL_P1_LEN(sl); + UBOUND(max, trash.size - trash.data - 3); + chunk_memcat(&trash, HTX_SL_P1_PTR(sl), max); + trash.area[trash.data++] = ' '; + + max = HTX_SL_P2_LEN(sl); + UBOUND(max, trash.size - trash.data - 2); + chunk_memcat(&trash, HTX_SL_P2_PTR(sl), max); + trash.area[trash.data++] = ' '; + + max = HTX_SL_P3_LEN(sl); + UBOUND(max, trash.size - trash.data - 1); + chunk_memcat(&trash, HTX_SL_P3_PTR(sl), max); + trash.area[trash.data++] = '\n'; + + DISGUISE(write(1, trash.area, trash.data)); +} + +/* + * Print a debug line with a header. + */ +static void http_debug_hdr(const char *dir, struct stream *s, const struct ist n, const struct ist v) +{ + struct session *sess = strm_sess(s); + int max; + + chunk_printf(&trash, "%08x:%s.%s[%04x:%04x]: ", s->uniq_id, s->be->id, + dir, + objt_conn(sess->origin) ? (unsigned short)__objt_conn(sess->origin)->handle.fd : -1, + sc_conn(s->scb) ? (unsigned short)(__sc_conn(s->scb))->handle.fd : -1); + + max = n.len; + UBOUND(max, trash.size - trash.data - 3); + chunk_memcat(&trash, n.ptr, max); + trash.area[trash.data++] = ':'; + trash.area[trash.data++] = ' '; + + max = v.len; + UBOUND(max, trash.size - trash.data - 1); + chunk_memcat(&trash, v.ptr, max); + trash.area[trash.data++] = '\n'; + + DISGUISE(write(1, trash.area, trash.data)); +} + +void http_txn_reset_req(struct http_txn *txn) +{ + txn->req.flags = 0; + txn->req.msg_state = HTTP_MSG_RQBEFORE; /* at the very beginning of the request */ +} + +void http_txn_reset_res(struct http_txn *txn) +{ + txn->rsp.flags = 0; + txn->rsp.msg_state = HTTP_MSG_RPBEFORE; /* at the very beginning of the response */ +} + +/* + * Create and initialize a new HTTP transaction for stream <s>. This should be + * used before processing any new request. It returns the transaction or NLULL + * on error. + */ +struct http_txn *http_create_txn(struct stream *s) +{ + struct http_txn *txn; + struct stconn *sc = s->scf; + + txn = pool_alloc(pool_head_http_txn); + if (!txn) + return NULL; + s->txn = txn; + + txn->meth = HTTP_METH_OTHER; + txn->flags = ((sc && sc_ep_test(sc, SE_FL_NOT_FIRST)) ? TX_NOT_FIRST : 0); + txn->status = -1; + txn->http_reply = NULL; + txn->l7_buffer = BUF_NULL; + write_u32(txn->cache_hash, 0); + + txn->cookie_first_date = 0; + txn->cookie_last_date = 0; + + txn->srv_cookie = NULL; + txn->cli_cookie = NULL; + txn->uri = NULL; + + http_txn_reset_req(txn); + http_txn_reset_res(txn); + + txn->req.chn = &s->req; + txn->rsp.chn = &s->res; + + txn->auth.method = HTTP_AUTH_UNKNOWN; + + /* here we don't want to re-initialize s->vars_txn and s->vars_reqres + * variable lists, because they were already initialized upon stream + * creation in stream_new(), and thus may already contain some variables + */ + + return txn; +} + +/* to be used at the end of a transaction */ +void http_destroy_txn(struct stream *s) +{ + struct http_txn *txn = s->txn; + + /* these ones will have been dynamically allocated */ + pool_free(pool_head_requri, txn->uri); + pool_free(pool_head_capture, txn->cli_cookie); + pool_free(pool_head_capture, txn->srv_cookie); + pool_free(pool_head_uniqueid, s->unique_id.ptr); + + s->unique_id = IST_NULL; + txn->uri = NULL; + txn->srv_cookie = NULL; + txn->cli_cookie = NULL; + + if (!LIST_ISEMPTY(&s->vars_txn.head)) + vars_prune(&s->vars_txn, s->sess, s); + if (!LIST_ISEMPTY(&s->vars_reqres.head)) + vars_prune(&s->vars_reqres, s->sess, s); + + b_free(&txn->l7_buffer); + + pool_free(pool_head_http_txn, txn); + s->txn = NULL; +} + + +DECLARE_POOL(pool_head_http_txn, "http_txn", sizeof(struct http_txn)); + +/* + * Local variables: + * c-indent-level: 8 + * c-basic-offset: 8 + * End: + */ |