blob: ddfb2222d4962a68ceee10033692d9b553c28d49 [file] [log] [blame]
Willy Tarreau79e57332018-10-02 16:01:16 +02001/*
2 * HTTP samples fetching
3 *
4 * Copyright 2000-2018 Willy Tarreau <w@1wt.eu>
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 */
12
13#include <sys/types.h>
14
15#include <ctype.h>
16#include <string.h>
17#include <time.h>
18
Willy Tarreaub2551052020-06-09 09:07:15 +020019#include <haproxy/api.h>
Willy Tarreaudfd3de82020-06-04 23:46:14 +020020#include <haproxy/arg.h>
Willy Tarreauac13aea2020-06-04 10:36:03 +020021#include <haproxy/auth.h>
Willy Tarreau8d366972020-05-27 16:10:29 +020022#include <haproxy/base64.h>
Willy Tarreauf1d32c42020-06-04 21:07:02 +020023#include <haproxy/channel.h>
Willy Tarreauc13ed532020-06-02 10:22:45 +020024#include <haproxy/chunk.h>
Willy Tarreau7ea393d2020-06-04 18:02:10 +020025#include <haproxy/connection.h>
Willy Tarreauf268ee82020-06-04 17:05:57 +020026#include <haproxy/global.h>
Willy Tarreau5413a872020-06-02 19:33:08 +020027#include <haproxy/h1.h>
Willy Tarreauc6fe8842020-06-04 09:00:02 +020028#include <haproxy/h1_htx.h>
Willy Tarreaucd72d8c2020-06-02 19:11:26 +020029#include <haproxy/http.h>
Willy Tarreauc2b1ff02020-06-04 21:21:03 +020030#include <haproxy/http_ana.h>
Willy Tarreau126ba3a2020-06-04 18:26:43 +020031#include <haproxy/http_fetch.h>
Willy Tarreau87735332020-06-04 09:08:41 +020032#include <haproxy/http_htx.h>
Willy Tarreau16f958c2020-06-03 08:44:35 +020033#include <haproxy/htx.h>
Willy Tarreau8efbdfb2020-06-04 11:29:21 +020034#include <haproxy/obj_type.h>
Willy Tarreaud0ef4392020-06-02 09:38:52 +020035#include <haproxy/pool.h>
Willy Tarreaue6ce10b2020-06-04 15:33:47 +020036#include <haproxy/sample.h>
Willy Tarreau5edca2f2022-05-27 09:25:10 +020037#include <haproxy/sc_strm.h>
Willy Tarreaudfd3de82020-06-04 23:46:14 +020038#include <haproxy/stream.h>
Willy Tarreau48fbcae2020-06-03 18:09:46 +020039#include <haproxy/tools.h>
Willy Tarreaud6788052020-05-27 15:59:00 +020040#include <haproxy/version.h>
Willy Tarreau79e57332018-10-02 16:01:16 +020041
Willy Tarreau79e57332018-10-02 16:01:16 +020042
43/* this struct is used between calls to smp_fetch_hdr() or smp_fetch_cookie() */
Christopher Fauletef453ed2018-10-24 21:39:27 +020044static THREAD_LOCAL struct http_hdr_ctx static_http_hdr_ctx;
Richard Russo458eafb2019-07-31 11:45:56 -070045/* this is used to convert raw connection buffers to htx */
46static THREAD_LOCAL struct buffer static_raw_htx_chunk;
47static THREAD_LOCAL char *static_raw_htx_buf;
Christopher Fauletef453ed2018-10-24 21:39:27 +020048
Christopher Faulet89dc4992019-04-17 12:02:59 +020049#define SMP_REQ_CHN(smp) (smp->strm ? &smp->strm->req : NULL)
50#define SMP_RES_CHN(smp) (smp->strm ? &smp->strm->res : NULL)
Willy Tarreau79e57332018-10-02 16:01:16 +020051
Richard Russo458eafb2019-07-31 11:45:56 -070052/* This function returns the static htx chunk, where raw connections get
53 * converted to HTX as needed for samplxsing.
54 */
55struct buffer *get_raw_htx_chunk(void)
56{
57 chunk_reset(&static_raw_htx_chunk);
58 return &static_raw_htx_chunk;
59}
60
61static int alloc_raw_htx_chunk_per_thread()
62{
63 static_raw_htx_buf = malloc(global.tune.bufsize);
64 if (!static_raw_htx_buf)
65 return 0;
66 chunk_init(&static_raw_htx_chunk, static_raw_htx_buf, global.tune.bufsize);
67 return 1;
68}
69
70static void free_raw_htx_chunk_per_thread()
71{
Willy Tarreau61cfdf42021-02-20 10:46:51 +010072 ha_free(&static_raw_htx_buf);
Richard Russo458eafb2019-07-31 11:45:56 -070073}
74
75REGISTER_PER_THREAD_ALLOC(alloc_raw_htx_chunk_per_thread);
76REGISTER_PER_THREAD_FREE(free_raw_htx_chunk_per_thread);
77
Willy Tarreau79e57332018-10-02 16:01:16 +020078/*
79 * Returns the data from Authorization header. Function may be called more
80 * than once so data is stored in txn->auth_data. When no header is found
81 * or auth method is unknown auth_method is set to HTTP_AUTH_WRONG to avoid
82 * searching again for something we are unable to find anyway. However, if
83 * the result if valid, the cache is not reused because we would risk to
84 * have the credentials overwritten by another stream in parallel.
Willy Tarreaueae83722020-04-29 11:52:51 +020085 * The caller is responsible for passing a sample with a valid stream/txn,
86 * and a valid htx.
Willy Tarreau79e57332018-10-02 16:01:16 +020087 */
88
Christopher Fauletcd761952019-07-15 13:58:29 +020089static int get_http_auth(struct sample *smp, struct htx *htx)
Willy Tarreau79e57332018-10-02 16:01:16 +020090{
Christopher Faulet311c7ea2018-10-24 21:41:55 +020091 struct stream *s = smp->strm;
Willy Tarreau79e57332018-10-02 16:01:16 +020092 struct http_txn *txn = s->txn;
Christopher Faulet6d1dd462019-07-15 14:36:03 +020093 struct http_hdr_ctx ctx = { .blk = NULL };
94 struct ist hdr;
Willy Tarreau79e57332018-10-02 16:01:16 +020095 struct buffer auth_method;
Christopher Faulet6d1dd462019-07-15 14:36:03 +020096 char *p;
Willy Tarreau79e57332018-10-02 16:01:16 +020097 int len;
98
99#ifdef DEBUG_AUTH
100 printf("Auth for stream %p: %d\n", s, txn->auth.method);
101#endif
Willy Tarreau79e57332018-10-02 16:01:16 +0200102 if (txn->auth.method == HTTP_AUTH_WRONG)
103 return 0;
104
105 txn->auth.method = HTTP_AUTH_WRONG;
106
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200107 if (txn->flags & TX_USE_PX_CONN)
108 hdr = ist("Proxy-Authorization");
109 else
110 hdr = ist("Authorization");
Willy Tarreau79e57332018-10-02 16:01:16 +0200111
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200112 ctx.blk = NULL;
113 if (!http_find_header(htx, hdr, &ctx, 0))
114 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200115
Willy Tarreau17254932020-09-02 07:08:47 +0200116 p = memchr(ctx.value.ptr, ' ', ctx.value.len);
117 if (!p || p == ctx.value.ptr) /* if no space was found or if the space is the first character */
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200118 return 0;
Willy Tarreau17254932020-09-02 07:08:47 +0200119 len = p - ctx.value.ptr;
Willy Tarreau79e57332018-10-02 16:01:16 +0200120
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200121 if (chunk_initlen(&auth_method, ctx.value.ptr, 0, len) != 1)
122 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200123
Remi Tricot-Le Breton68c4eae2021-10-29 15:25:18 +0200124 /* According to RFC7235, there could be multiple spaces between the
125 * scheme and its value, we must skip all of them.
126 */
127 while (p < istend(ctx.value) && *p == ' ')
128 ++p;
129
130 chunk_initlen(&txn->auth.method_data, p, 0, istend(ctx.value) - p);
Willy Tarreau79e57332018-10-02 16:01:16 +0200131
132 if (!strncasecmp("Basic", auth_method.area, auth_method.data)) {
133 struct buffer *http_auth = get_trash_chunk();
134
135 len = base64dec(txn->auth.method_data.area,
136 txn->auth.method_data.data,
137 http_auth->area, global.tune.bufsize - 1);
138
139 if (len < 0)
140 return 0;
141
142
143 http_auth->area[len] = '\0';
144
145 p = strchr(http_auth->area, ':');
146
147 if (!p)
148 return 0;
149
150 txn->auth.user = http_auth->area;
151 *p = '\0';
152 txn->auth.pass = p+1;
153
154 txn->auth.method = HTTP_AUTH_BASIC;
155 return 1;
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +0200156 } else if (!strncasecmp("Bearer", auth_method.area, auth_method.data)) {
157 txn->auth.method = HTTP_AUTH_BEARER;
158 return 1;
Willy Tarreau79e57332018-10-02 16:01:16 +0200159 }
160
161 return 0;
162}
163
164/* This function ensures that the prerequisites for an L7 fetch are ready,
165 * which means that a request or response is ready. If some data is missing,
166 * a parsing attempt is made. This is useful in TCP-based ACLs which are able
Christopher Faulet5ec8bcb2019-04-17 12:04:12 +0200167 * to extract data from L7. If <vol> is non-null during a prefetch, another
168 * test is made to ensure the required information is not gone.
Christopher Fauletef453ed2018-10-24 21:39:27 +0200169 *
170 * The function returns :
171 * NULL with SMP_F_MAY_CHANGE in the sample flags if some data is missing to
172 * decide whether or not an HTTP message is present ;
173 * NULL if the requested data cannot be fetched or if it is certain that
Willy Tarreaueae83722020-04-29 11:52:51 +0200174 * we'll never have any HTTP message there; this includes null strm or chn.
Willy Tarreaua6d98792020-08-12 14:04:52 +0200175 * NULL if the sample's direction does not match the channel's (i.e. the
176 * function was asked to work on the wrong channel)
Christopher Fauletef453ed2018-10-24 21:39:27 +0200177 * The HTX message if ready
178 */
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200179struct htx *smp_prefetch_htx(struct sample *smp, struct channel *chn, struct check *check, int vol)
Christopher Fauletef453ed2018-10-24 21:39:27 +0200180{
Christopher Fauletef453ed2018-10-24 21:39:27 +0200181 struct stream *s = smp->strm;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200182 struct http_txn *txn = NULL;
183 struct htx *htx = NULL;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200184 struct http_msg *msg;
Christopher Fauletf1ba18d2018-11-26 21:37:08 +0100185 struct htx_sl *sl;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200186
Willy Tarreaua6d98792020-08-12 14:04:52 +0200187 if (chn &&
188 (((smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ && (chn->flags & CF_ISRESP)) ||
189 ((smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_RES && !(chn->flags & CF_ISRESP))))
190 return 0;
191
Christopher Fauletef453ed2018-10-24 21:39:27 +0200192 /* Note: it is possible that <s> is NULL when called before stream
193 * initialization (eg: tcp-request connection), so this function is the
194 * one responsible for guarding against this case for all HTTP users.
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200195 *
196 * In the health check context, the stream and the channel must be NULL
197 * and <check> must be set. In this case, only the input buffer,
198 * corresponding to the response, is considered. It is the caller
199 * responsibility to provide <check>.
Christopher Fauletef453ed2018-10-24 21:39:27 +0200200 */
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200201 BUG_ON(check && (s || chn));
202 if (!s || !chn) {
203 if (check) {
204 htx = htxbuf(&check->bi);
205
206 /* Analyse not yet started */
207 if (htx_is_empty(htx) || htx->first == -1)
208 return NULL;
209
210 sl = http_get_stline(htx);
211 if (vol && !sl) {
212 /* The start-line was already forwarded, it is too late to fetch anything */
213 return NULL;
214 }
215 goto end;
216 }
217
Christopher Fauletef453ed2018-10-24 21:39:27 +0200218 return NULL;
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200219 }
Christopher Fauletef453ed2018-10-24 21:39:27 +0200220
Christopher Faulet75f619a2021-03-08 19:12:58 +0100221 if (!s->txn && !http_create_txn(s))
222 return NULL;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200223 txn = s->txn;
224 msg = (!(chn->flags & CF_ISRESP) ? &txn->req : &txn->rsp);
Christopher Fauletef453ed2018-10-24 21:39:27 +0200225
Christopher Fauleteca88542019-04-03 10:12:42 +0200226 if (IS_HTX_STRM(s)) {
Christopher Faulet89dc4992019-04-17 12:02:59 +0200227 htx = htxbuf(&chn->buf);
Christopher Fauletef453ed2018-10-24 21:39:27 +0200228
Christopher Faulet89dc4992019-04-17 12:02:59 +0200229 if (msg->msg_state == HTTP_MSG_ERROR || (htx->flags & HTX_FL_PARSING_ERROR))
230 return NULL;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200231
Christopher Faulet89dc4992019-04-17 12:02:59 +0200232 if (msg->msg_state < HTTP_MSG_BODY) {
233 /* Analyse not yet started */
Christopher Faulet29f17582019-05-23 11:03:26 +0200234 if (htx_is_empty(htx) || htx->first == -1) {
Christopher Fauletef453ed2018-10-24 21:39:27 +0200235 /* Parsing is done by the mux, just wait */
236 smp->flags |= SMP_F_MAY_CHANGE;
237 return NULL;
238 }
239 }
Christopher Faulet297fbb42019-05-13 14:41:27 +0200240 sl = http_get_stline(htx);
Christopher Faulet5ec8bcb2019-04-17 12:04:12 +0200241 if (vol && !sl) {
Christopher Faulet89dc4992019-04-17 12:02:59 +0200242 /* The start-line was already forwarded, it is too late to fetch anything */
243 return NULL;
244 }
Christopher Fauletef453ed2018-10-24 21:39:27 +0200245 }
Christopher Fauleteca88542019-04-03 10:12:42 +0200246 else { /* RAW mode */
Christopher Faulet89dc4992019-04-17 12:02:59 +0200247 struct buffer *buf;
248 struct h1m h1m;
Christopher Faulete4ab11b2019-06-11 15:05:37 +0200249 struct http_hdr hdrs[global.tune.max_http_hdr];
Christopher Faulet89dc4992019-04-17 12:02:59 +0200250 union h1_sl h1sl;
251 unsigned int flags = HTX_FL_NONE;
252 int ret;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200253
Christopher Faulet89dc4992019-04-17 12:02:59 +0200254 /* no HTTP fetch on the response in TCP mode */
255 if (chn->flags & CF_ISRESP)
256 return NULL;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200257
Christopher Faulet89dc4992019-04-17 12:02:59 +0200258 /* Now we are working on the request only */
259 buf = &chn->buf;
260 if (b_head(buf) + b_data(buf) > b_wrap(buf))
261 b_slow_realign(buf, trash.area, 0);
Christopher Fauletef453ed2018-10-24 21:39:27 +0200262
Christopher Faulet89dc4992019-04-17 12:02:59 +0200263 h1m_init_req(&h1m);
264 ret = h1_headers_to_hdr_list(b_head(buf), b_stop(buf),
265 hdrs, sizeof(hdrs)/sizeof(hdrs[0]), &h1m, &h1sl);
266 if (ret <= 0) {
267 /* Invalid or too big*/
268 if (ret < 0 || channel_full(&s->req, global.tune.maxrewrite))
Christopher Fauletef453ed2018-10-24 21:39:27 +0200269 return NULL;
Christopher Fauletf1ba18d2018-11-26 21:37:08 +0100270
Christopher Faulet89dc4992019-04-17 12:02:59 +0200271 /* wait for a full request */
272 smp->flags |= SMP_F_MAY_CHANGE;
273 return NULL;
274 }
Christopher Fauletf1ba18d2018-11-26 21:37:08 +0100275
Ilya Shipitsin46a030c2020-07-05 16:36:08 +0500276 /* OK we just got a valid HTTP message. We have to convert it
Christopher Faulet89dc4992019-04-17 12:02:59 +0200277 * into an HTX message.
278 */
279 if (unlikely(h1sl.rq.v.len == 0)) {
280 /* try to convert HTTP/0.9 requests to HTTP/1.0 */
281 if (h1sl.rq.meth != HTTP_METH_GET || !h1sl.rq.u.len)
Christopher Fauletef453ed2018-10-24 21:39:27 +0200282 return NULL;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200283 h1sl.rq.v = ist("HTTP/1.0");
Christopher Fauletef453ed2018-10-24 21:39:27 +0200284 }
Christopher Faulet89dc4992019-04-17 12:02:59 +0200285
286 /* Set HTX start-line flags */
287 if (h1m.flags & H1_MF_VER_11)
288 flags |= HTX_SL_F_VER_11;
289 if (h1m.flags & H1_MF_XFER_ENC)
290 flags |= HTX_SL_F_XFER_ENC;
291 flags |= HTX_SL_F_XFER_LEN;
292 if (h1m.flags & H1_MF_CHNK)
293 flags |= HTX_SL_F_CHNK;
294 else if (h1m.flags & H1_MF_CLEN)
295 flags |= HTX_SL_F_CLEN;
296
Richard Russo458eafb2019-07-31 11:45:56 -0700297 htx = htx_from_buf(get_raw_htx_chunk());
Christopher Faulet89dc4992019-04-17 12:02:59 +0200298 sl = htx_add_stline(htx, HTX_BLK_REQ_SL, flags, h1sl.rq.m, h1sl.rq.u, h1sl.rq.v);
299 if (!sl || !htx_add_all_headers(htx, hdrs))
Christopher Fauletef453ed2018-10-24 21:39:27 +0200300 return NULL;
Willy Tarreauce9bbf52019-05-13 08:32:31 +0200301 sl->info.req.meth = h1sl.rq.meth;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200302 }
303
304 /* OK we just got a valid HTTP message. If not already done by
305 * HTTP analyzers, we have some minor preparation to perform so
306 * that further checks can rely on HTTP tests.
307 */
308 if (sl && msg->msg_state < HTTP_MSG_BODY) {
309 if (!(chn->flags & CF_ISRESP)) {
310 txn->meth = sl->info.req.meth;
311 if (txn->meth == HTTP_METH_GET || txn->meth == HTTP_METH_HEAD)
312 s->flags |= SF_REDIRECTABLE;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200313 }
Christopher Faulet89dc4992019-04-17 12:02:59 +0200314 else
315 txn->status = sl->info.res.status;
316 if (sl->flags & HTX_SL_F_VER_11)
317 msg->flags |= HTTP_MSGF_VER_11;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200318 }
319
320 /* everything's OK */
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200321 end:
Christopher Fauletef453ed2018-10-24 21:39:27 +0200322 return htx;
323}
324
Willy Tarreau79e57332018-10-02 16:01:16 +0200325/* This function fetches the method of current HTTP request and stores
326 * it in the global pattern struct as a chunk. There are two possibilities :
327 * - if the method is known (not HTTP_METH_OTHER), its identifier is stored
328 * in <len> and <ptr> is NULL ;
329 * - if the method is unknown (HTTP_METH_OTHER), <ptr> points to the text and
330 * <len> to its length.
331 * This is intended to be used with pat_match_meth() only.
332 */
333static int smp_fetch_meth(const struct arg *args, struct sample *smp, const char *kw, void *private)
334{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200335 struct channel *chn = SMP_REQ_CHN(smp);
Willy Tarreau79e57332018-10-02 16:01:16 +0200336 struct http_txn *txn;
Christopher Fauletdbbdb252022-06-22 17:16:41 +0200337 struct htx *htx;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200338 int meth;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200339
Christopher Faulet12f6dbb2022-07-06 17:53:02 +0200340 txn = (smp->strm ? smp->strm->txn : NULL);
Willy Tarreaua6d98792020-08-12 14:04:52 +0200341 if (!txn)
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200342 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200343
Willy Tarreaua88e8bf2022-07-10 13:13:52 +0200344 meth = txn->meth;
345 if (meth == HTTP_METH_OTHER) {
Christopher Fauletdbbdb252022-06-22 17:16:41 +0200346 htx = smp_prefetch_htx(smp, chn, NULL, 1);
347 if (!htx)
348 return 0;
349 }
350
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200351 smp->data.type = SMP_T_METH;
352 smp->data.u.meth.meth = meth;
353 if (meth == HTTP_METH_OTHER) {
354 struct htx_sl *sl;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200355
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200356 sl = http_get_stline(htx);
357 smp->flags |= SMP_F_CONST;
358 smp->data.u.meth.str.area = HTX_SL_REQ_MPTR(sl);
359 smp->data.u.meth.str.data = HTX_SL_REQ_MLEN(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +0200360 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200361 smp->flags |= SMP_F_VOL_1ST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200362 return 1;
363}
364
365static int smp_fetch_rqver(const struct arg *args, struct sample *smp, const char *kw, void *private)
366{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200367 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200368 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200369 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200370 char *ptr;
371 int len;
372
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200373 if (!htx)
374 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200375
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200376 sl = http_get_stline(htx);
377 len = HTX_SL_REQ_VLEN(sl);
378 ptr = HTX_SL_REQ_VPTR(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +0200379
380 while ((len-- > 0) && (*ptr++ != '/'));
381 if (len <= 0)
382 return 0;
383
384 smp->data.type = SMP_T_STR;
385 smp->data.u.str.area = ptr;
386 smp->data.u.str.data = len;
387
388 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
389 return 1;
390}
391
392static int smp_fetch_stver(const struct arg *args, struct sample *smp, const char *kw, void *private)
393{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200394 struct channel *chn = SMP_RES_CHN(smp);
Christopher Fauletf98e6262020-05-06 09:42:04 +0200395 struct check *check = objt_check(smp->sess->origin);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200396 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200397 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200398 char *ptr;
399 int len;
400
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200401 if (!htx)
402 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200403
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200404 sl = http_get_stline(htx);
405 len = HTX_SL_RES_VLEN(sl);
406 ptr = HTX_SL_RES_VPTR(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +0200407
408 while ((len-- > 0) && (*ptr++ != '/'));
409 if (len <= 0)
410 return 0;
411
412 smp->data.type = SMP_T_STR;
413 smp->data.u.str.area = ptr;
414 smp->data.u.str.data = len;
415
416 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
417 return 1;
418}
419
420/* 3. Check on Status Code. We manipulate integers here. */
421static int smp_fetch_stcode(const struct arg *args, struct sample *smp, const char *kw, void *private)
422{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200423 struct channel *chn = SMP_RES_CHN(smp);
Christopher Fauletf98e6262020-05-06 09:42:04 +0200424 struct check *check = objt_check(smp->sess->origin);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200425 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200426 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200427 char *ptr;
428 int len;
429
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200430 if (!htx)
431 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200432
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200433 sl = http_get_stline(htx);
434 len = HTX_SL_RES_CLEN(sl);
435 ptr = HTX_SL_RES_CPTR(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +0200436
437 smp->data.type = SMP_T_SINT;
438 smp->data.u.sint = __strl2ui(ptr, len);
439 smp->flags = SMP_F_VOL_1ST;
440 return 1;
441}
442
443static int smp_fetch_uniqueid(const struct arg *args, struct sample *smp, const char *kw, void *private)
444{
Tim Duesterhusa17e6622020-03-05 20:19:02 +0100445 struct ist unique_id;
Tim Duesterhus2825b4b2020-02-28 15:13:34 +0100446
Willy Tarreau79e57332018-10-02 16:01:16 +0200447 if (LIST_ISEMPTY(&smp->sess->fe->format_unique_id))
448 return 0;
449
Willy Tarreaua1062a42020-04-29 11:50:38 +0200450 if (!smp->strm)
451 return 0;
452
Tim Duesterhusa17e6622020-03-05 20:19:02 +0100453 unique_id = stream_generate_unique_id(smp->strm, &smp->sess->fe->format_unique_id);
454 if (!isttest(unique_id))
Tim Duesterhus2825b4b2020-02-28 15:13:34 +0100455 return 0;
456
Tim Duesterhusa17e6622020-03-05 20:19:02 +0100457 smp->data.u.str.area = smp->strm->unique_id.ptr;
458 smp->data.u.str.data = smp->strm->unique_id.len;
Tim Duesterhus2825b4b2020-02-28 15:13:34 +0100459 smp->data.type = SMP_T_STR;
Willy Tarreau79e57332018-10-02 16:01:16 +0200460 smp->flags = SMP_F_CONST;
461 return 1;
462}
463
464/* Returns a string block containing all headers including the
Joseph Herlant942eea32018-11-15 13:57:22 -0800465 * empty line which separes headers from the body. This is useful
466 * for some headers analysis.
Willy Tarreau79e57332018-10-02 16:01:16 +0200467 */
468static int smp_fetch_hdrs(const struct arg *args, struct sample *smp, const char *kw, void *private)
469{
Christopher Faulete596d182020-05-05 17:46:34 +0200470 /* possible keywords: req.hdrs, res.hdrs */
471 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200472 struct check *check = ((kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200473 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200474 struct buffer *temp;
475 int32_t pos;
Willy Tarreau79e57332018-10-02 16:01:16 +0200476
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200477 if (!htx)
478 return 0;
479 temp = get_trash_chunk();
480 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
481 struct htx_blk *blk = htx_get_blk(htx, pos);
482 enum htx_blk_type type = htx_get_blk_type(blk);
Willy Tarreau79e57332018-10-02 16:01:16 +0200483
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200484 if (type == HTX_BLK_HDR) {
485 struct ist n = htx_get_blk_name(htx, blk);
486 struct ist v = htx_get_blk_value(htx, blk);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200487
Christopher Faulet53a899b2019-10-08 16:38:42 +0200488 if (!h1_format_htx_hdr(n, v, temp))
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200489 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200490 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200491 else if (type == HTX_BLK_EOH) {
492 if (!chunk_memcat(temp, "\r\n", 2))
493 return 0;
494 break;
495 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200496 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200497 smp->data.type = SMP_T_STR;
498 smp->data.u.str = *temp;
Willy Tarreau79e57332018-10-02 16:01:16 +0200499 return 1;
500}
501
502/* Returns the header request in a length/value encoded format.
503 * This is useful for exchanges with the SPOE.
504 *
505 * A "length value" is a multibyte code encoding numbers. It uses the
506 * SPOE format. The encoding is the following:
507 *
508 * Each couple "header name" / "header value" is composed
509 * like this:
510 * "length value" "header name bytes"
511 * "length value" "header value bytes"
512 * When the last header is reached, the header name and the header
513 * value are empty. Their length are 0
514 */
515static int smp_fetch_hdrs_bin(const struct arg *args, struct sample *smp, const char *kw, void *private)
516{
Christopher Faulete596d182020-05-05 17:46:34 +0200517 /* possible keywords: req.hdrs_bin, res.hdrs_bin */
518 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200519 struct check *check = ((kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200520 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200521 struct buffer *temp;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200522 char *p, *end;
523 int32_t pos;
524 int ret;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200525
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200526 if (!htx)
527 return 0;
528 temp = get_trash_chunk();
529 p = temp->area;
530 end = temp->area + temp->size;
531 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
532 struct htx_blk *blk = htx_get_blk(htx, pos);
533 enum htx_blk_type type = htx_get_blk_type(blk);
534 struct ist n, v;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200535
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200536 if (type == HTX_BLK_HDR) {
537 n = htx_get_blk_name(htx,blk);
538 v = htx_get_blk_value(htx, blk);
Willy Tarreau79e57332018-10-02 16:01:16 +0200539
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200540 /* encode the header name. */
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200541 ret = encode_varint(n.len, &p, end);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200542 if (ret == -1)
543 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200544 if (p + n.len > end)
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200545 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200546 memcpy(p, n.ptr, n.len);
547 p += n.len;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200548
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200549 /* encode the header value. */
550 ret = encode_varint(v.len, &p, end);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200551 if (ret == -1)
552 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200553 if (p + v.len > end)
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200554 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200555 memcpy(p, v.ptr, v.len);
556 p += v.len;
Willy Tarreau79e57332018-10-02 16:01:16 +0200557
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200558 }
559 else if (type == HTX_BLK_EOH) {
560 /* encode the end of the header list with empty
561 * header name and header value.
562 */
563 ret = encode_varint(0, &p, end);
564 if (ret == -1)
565 return 0;
566 ret = encode_varint(0, &p, end);
567 if (ret == -1)
568 return 0;
569 break;
570 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200571 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200572
573 /* Initialise sample data which will be filled. */
574 smp->data.type = SMP_T_BIN;
575 smp->data.u.str.area = temp->area;
576 smp->data.u.str.data = p - temp->area;
577 smp->data.u.str.size = temp->size;
Willy Tarreau79e57332018-10-02 16:01:16 +0200578 return 1;
579}
580
581/* returns the longest available part of the body. This requires that the body
582 * has been waited for using http-buffer-request.
583 */
584static int smp_fetch_body(const struct arg *args, struct sample *smp, const char *kw, void *private)
585{
Christopher Faulete596d182020-05-05 17:46:34 +0200586 /* possible keywords: req.body, res.body */
587 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200588 struct check *check = ((kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200589 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Willy Tarreau79e57332018-10-02 16:01:16 +0200590 struct buffer *temp;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200591 int32_t pos;
Christopher Fauleta9ffc412020-11-25 08:08:08 +0100592 int finished = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200593
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200594 if (!htx)
595 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200596
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200597 temp = get_trash_chunk();
598 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
599 struct htx_blk *blk = htx_get_blk(htx, pos);
600 enum htx_blk_type type = htx_get_blk_type(blk);
Willy Tarreau79e57332018-10-02 16:01:16 +0200601
Christopher Fauletd1ac2b92020-12-02 19:12:22 +0100602 if (type == HTX_BLK_TLR || type == HTX_BLK_EOT) {
Christopher Fauleta9ffc412020-11-25 08:08:08 +0100603 finished = 1;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200604 break;
Christopher Fauleta9ffc412020-11-25 08:08:08 +0100605 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200606 if (type == HTX_BLK_DATA) {
Christopher Faulet53a899b2019-10-08 16:38:42 +0200607 if (!h1_format_htx_data(htx_get_blk_value(htx, blk), temp, 0))
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200608 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200609 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200610 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200611
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200612 smp->data.type = SMP_T_BIN;
613 smp->data.u.str = *temp;
614 smp->flags = SMP_F_VOL_TEST;
Willy Tarreau9dc92b22020-06-15 18:01:10 +0200615
Christopher Fauleta9ffc412020-11-25 08:08:08 +0100616 if (!finished && (check || (chn && !channel_full(chn, global.tune.maxrewrite) &&
617 !(chn->flags & (CF_EOI|CF_SHUTR|CF_READ_ERROR)))))
Willy Tarreau9dc92b22020-06-15 18:01:10 +0200618 smp->flags |= SMP_F_MAY_CHANGE;
619
Willy Tarreau79e57332018-10-02 16:01:16 +0200620 return 1;
621}
622
623
624/* returns the available length of the body. This requires that the body
625 * has been waited for using http-buffer-request.
626 */
627static int smp_fetch_body_len(const struct arg *args, struct sample *smp, const char *kw, void *private)
628{
Christopher Faulete596d182020-05-05 17:46:34 +0200629 /* possible keywords: req.body_len, res.body_len */
630 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200631 struct check *check = ((kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200632 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200633 int32_t pos;
634 unsigned long long len = 0;
Christopher Fauletc16317d2018-12-12 14:11:22 +0100635
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200636 if (!htx)
637 return 0;
Christopher Fauletc16317d2018-12-12 14:11:22 +0100638
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200639 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
640 struct htx_blk *blk = htx_get_blk(htx, pos);
641 enum htx_blk_type type = htx_get_blk_type(blk);
Christopher Fauletc16317d2018-12-12 14:11:22 +0100642
Christopher Fauletd1ac2b92020-12-02 19:12:22 +0100643 if (type == HTX_BLK_TLR || type == HTX_BLK_EOT)
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200644 break;
645 if (type == HTX_BLK_DATA)
646 len += htx_get_blksz(blk);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200647 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200648
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200649 smp->data.type = SMP_T_SINT;
650 smp->data.u.sint = len;
651 smp->flags = SMP_F_VOL_TEST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200652 return 1;
653}
654
655
656/* returns the advertised length of the body, or the advertised size of the
657 * chunks available in the buffer. This requires that the body has been waited
658 * for using http-buffer-request.
659 */
660static int smp_fetch_body_size(const struct arg *args, struct sample *smp, const char *kw, void *private)
661{
Christopher Faulete596d182020-05-05 17:46:34 +0200662 /* possible keywords: req.body_size, res.body_size */
663 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200664 struct check *check = ((kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200665 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200666 int32_t pos;
667 unsigned long long len = 0;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200668
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200669 if (!htx)
670 return 0;
Christopher Fauletc16317d2018-12-12 14:11:22 +0100671
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200672 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
673 struct htx_blk *blk = htx_get_blk(htx, pos);
674 enum htx_blk_type type = htx_get_blk_type(blk);
Christopher Fauletc16317d2018-12-12 14:11:22 +0100675
Christopher Fauletd1ac2b92020-12-02 19:12:22 +0100676 if (type == HTX_BLK_TLR || type == HTX_BLK_EOT)
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200677 break;
678 if (type == HTX_BLK_DATA)
679 len += htx_get_blksz(blk);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200680 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200681 if (htx->extra != ULLONG_MAX)
682 len += htx->extra;
Willy Tarreau79e57332018-10-02 16:01:16 +0200683
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200684 smp->data.type = SMP_T_SINT;
685 smp->data.u.sint = len;
686 smp->flags = SMP_F_VOL_TEST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200687 return 1;
688}
689
690
691/* 4. Check on URL/URI. A pointer to the URI is stored. */
692static int smp_fetch_url(const struct arg *args, struct sample *smp, const char *kw, void *private)
693{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200694 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200695 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200696 struct htx_sl *sl;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200697
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200698 if (!htx)
699 return 0;
700 sl = http_get_stline(htx);
701 smp->data.type = SMP_T_STR;
702 smp->data.u.str.area = HTX_SL_REQ_UPTR(sl);
703 smp->data.u.str.data = HTX_SL_REQ_ULEN(sl);
704 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200705 return 1;
706}
707
708static int smp_fetch_url_ip(const struct arg *args, struct sample *smp, const char *kw, void *private)
709{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200710 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200711 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200712 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200713 struct sockaddr_storage addr;
714
Amaury Denoyellec89d5332021-05-10 11:23:34 +0200715 memset(&addr, 0, sizeof(addr));
716
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200717 if (!htx)
718 return 0;
719 sl = http_get_stline(htx);
Amaury Denoyellec89d5332021-05-10 11:23:34 +0200720 if (url2sa(HTX_SL_REQ_UPTR(sl), HTX_SL_REQ_ULEN(sl), &addr, NULL) < 0)
721 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200722
Willy Tarreau48584642021-05-09 10:32:54 +0200723 if (addr.ss_family != AF_INET)
Willy Tarreau79e57332018-10-02 16:01:16 +0200724 return 0;
725
726 smp->data.type = SMP_T_IPV4;
727 smp->data.u.ipv4 = ((struct sockaddr_in *)&addr)->sin_addr;
728 smp->flags = 0;
729 return 1;
730}
731
732static int smp_fetch_url_port(const struct arg *args, struct sample *smp, const char *kw, void *private)
733{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200734 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +0200735 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200736 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200737 struct sockaddr_storage addr;
738
Amaury Denoyellec89d5332021-05-10 11:23:34 +0200739 memset(&addr, 0, sizeof(addr));
740
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200741 if (!htx)
742 return 0;
743 sl = http_get_stline(htx);
Amaury Denoyellec89d5332021-05-10 11:23:34 +0200744 if (url2sa(HTX_SL_REQ_UPTR(sl), HTX_SL_REQ_ULEN(sl), &addr, NULL) < 0)
745 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200746
Willy Tarreau48584642021-05-09 10:32:54 +0200747 if (addr.ss_family != AF_INET)
Willy Tarreau79e57332018-10-02 16:01:16 +0200748 return 0;
749
750 smp->data.type = SMP_T_SINT;
Willy Tarreau48584642021-05-09 10:32:54 +0200751 smp->data.u.sint = get_host_port(&addr);
Willy Tarreau79e57332018-10-02 16:01:16 +0200752 smp->flags = 0;
753 return 1;
754}
755
756/* Fetch an HTTP header. A pointer to the beginning of the value is returned.
757 * Accepts an optional argument of type string containing the header field name,
758 * and an optional argument of type signed or unsigned integer to request an
759 * explicit occurrence of the header. Note that in the event of a missing name,
760 * headers are considered from the first one. It does not stop on commas and
761 * returns full lines instead (useful for User-Agent or Date for example).
762 */
763static int smp_fetch_fhdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
764{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200765 /* possible keywords: req.fhdr, res.fhdr */
766 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200767 struct check *check = ((kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200768 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200769 struct http_hdr_ctx *ctx = smp->ctx.a[0];
770 struct ist name;
Willy Tarreau79e57332018-10-02 16:01:16 +0200771 int occ = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200772
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200773 if (!ctx) {
774 /* first call */
775 ctx = &static_http_hdr_ctx;
776 ctx->blk = NULL;
777 smp->ctx.a[0] = ctx;
778 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200779
Christopher Faulet623af932021-01-29 11:22:15 +0100780 if (args[0].type != ARGT_STR)
781 return 0;
Tim Duesterhus92c696e2021-02-28 16:11:36 +0100782 name = ist2(args[0].data.str.area, args[0].data.str.data);
Willy Tarreau79e57332018-10-02 16:01:16 +0200783
Christopher Faulet623af932021-01-29 11:22:15 +0100784 if (args[1].type == ARGT_SINT)
785 occ = args[1].data.sint;
Willy Tarreau79e57332018-10-02 16:01:16 +0200786
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200787 if (!htx)
788 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200789
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200790 if (ctx && !(smp->flags & SMP_F_NOT_LAST))
791 /* search for header from the beginning */
792 ctx->blk = NULL;
Willy Tarreau79e57332018-10-02 16:01:16 +0200793
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200794 if (!occ && !(smp->opt & SMP_OPT_ITERATE))
795 /* no explicit occurrence and single fetch => last header by default */
796 occ = -1;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200797
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200798 if (!occ)
799 /* prepare to report multiple occurrences for ACL fetches */
800 smp->flags |= SMP_F_NOT_LAST;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200801
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200802 smp->data.type = SMP_T_STR;
803 smp->flags |= SMP_F_VOL_HDR | SMP_F_CONST;
804 if (http_get_htx_fhdr(htx, name, occ, ctx, &smp->data.u.str.area, &smp->data.u.str.data))
805 return 1;
Willy Tarreau79e57332018-10-02 16:01:16 +0200806 smp->flags &= ~SMP_F_NOT_LAST;
807 return 0;
808}
809
810/* 6. Check on HTTP header count. The number of occurrences is returned.
811 * Accepts exactly 1 argument of type string. It does not stop on commas and
812 * returns full lines instead (useful for User-Agent or Date for example).
813 */
814static int smp_fetch_fhdr_cnt(const struct arg *args, struct sample *smp, const char *kw, void *private)
815{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200816 /* possible keywords: req.fhdr_cnt, res.fhdr_cnt */
817 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200818 struct check *check = ((kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200819 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200820 struct http_hdr_ctx ctx;
821 struct ist name;
Willy Tarreau79e57332018-10-02 16:01:16 +0200822 int cnt;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200823
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200824 if (!htx)
825 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200826
Christopher Faulet623af932021-01-29 11:22:15 +0100827 if (args->type == ARGT_STR) {
Tim Duesterhus92c696e2021-02-28 16:11:36 +0100828 name = ist2(args->data.str.area, args->data.str.data);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200829 } else {
Tim Duesterhus68a088d2021-02-28 16:11:37 +0100830 name = IST_NULL;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200831 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200832
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200833 ctx.blk = NULL;
834 cnt = 0;
835 while (http_find_header(htx, name, &ctx, 1))
836 cnt++;
Willy Tarreau79e57332018-10-02 16:01:16 +0200837 smp->data.type = SMP_T_SINT;
838 smp->data.u.sint = cnt;
839 smp->flags = SMP_F_VOL_HDR;
840 return 1;
841}
842
843static int smp_fetch_hdr_names(const struct arg *args, struct sample *smp, const char *kw, void *private)
844{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200845 /* possible keywords: req.hdr_names, res.hdr_names */
846 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200847 struct check *check = ((kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200848 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Willy Tarreau79e57332018-10-02 16:01:16 +0200849 struct buffer *temp;
850 char del = ',';
851
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200852 int32_t pos;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200853
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200854 if (!htx)
855 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200856
Christopher Faulet623af932021-01-29 11:22:15 +0100857 if (args->type == ARGT_STR)
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200858 del = *args[0].data.str.area;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200859
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200860 temp = get_trash_chunk();
861 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
862 struct htx_blk *blk = htx_get_blk(htx, pos);
863 enum htx_blk_type type = htx_get_blk_type(blk);
864 struct ist n;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200865
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200866 if (type == HTX_BLK_EOH)
867 break;
868 if (type != HTX_BLK_HDR)
869 continue;
870 n = htx_get_blk_name(htx, blk);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200871
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200872 if (temp->data)
873 temp->area[temp->data++] = del;
Tim Duesterhus9f7ed8a2021-11-08 09:05:04 +0100874 chunk_istcat(temp, n);
Willy Tarreau79e57332018-10-02 16:01:16 +0200875 }
876
877 smp->data.type = SMP_T_STR;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200878 smp->data.u.str = *temp;
Willy Tarreau79e57332018-10-02 16:01:16 +0200879 smp->flags = SMP_F_VOL_HDR;
880 return 1;
881}
882
883/* Fetch an HTTP header. A pointer to the beginning of the value is returned.
884 * Accepts an optional argument of type string containing the header field name,
885 * and an optional argument of type signed or unsigned integer to request an
886 * explicit occurrence of the header. Note that in the event of a missing name,
887 * headers are considered from the first one.
888 */
889static int smp_fetch_hdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
890{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200891 /* possible keywords: req.hdr / hdr, res.hdr / shdr */
892 struct channel *chn = ((kw[0] == 'h' || kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200893 struct check *check = ((kw[0] == 's' || kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200894 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200895 struct http_hdr_ctx *ctx = smp->ctx.a[0];
896 struct ist name;
Willy Tarreau79e57332018-10-02 16:01:16 +0200897 int occ = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200898
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200899 if (!ctx) {
900 /* first call */
901 ctx = &static_http_hdr_ctx;
902 ctx->blk = NULL;
903 smp->ctx.a[0] = ctx;
904 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200905
Christopher Faulet623af932021-01-29 11:22:15 +0100906 if (args[0].type != ARGT_STR)
907 return 0;
Tim Duesterhus92c696e2021-02-28 16:11:36 +0100908 name = ist2(args[0].data.str.area, args[0].data.str.data);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200909
Christopher Faulet623af932021-01-29 11:22:15 +0100910 if (args[1].type == ARGT_SINT)
911 occ = args[1].data.sint;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200912
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200913 if (!htx)
914 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200915
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200916 if (ctx && !(smp->flags & SMP_F_NOT_LAST))
917 /* search for header from the beginning */
918 ctx->blk = NULL;
Willy Tarreau79e57332018-10-02 16:01:16 +0200919
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200920 if (!occ && !(smp->opt & SMP_OPT_ITERATE))
921 /* no explicit occurrence and single fetch => last header by default */
922 occ = -1;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200923
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200924 if (!occ)
925 /* prepare to report multiple occurrences for ACL fetches */
926 smp->flags |= SMP_F_NOT_LAST;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200927
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200928 smp->data.type = SMP_T_STR;
929 smp->flags |= SMP_F_VOL_HDR | SMP_F_CONST;
930 if (http_get_htx_hdr(htx, name, occ, ctx, &smp->data.u.str.area, &smp->data.u.str.data))
931 return 1;
Willy Tarreau79e57332018-10-02 16:01:16 +0200932
933 smp->flags &= ~SMP_F_NOT_LAST;
934 return 0;
935}
936
Christopher Fauletc1f40dd2019-05-16 10:07:30 +0200937/* Same than smp_fetch_hdr() but only relies on the sample direction to choose
938 * the right channel. So instead of duplicating the code, we just change the
939 * keyword and then fallback on smp_fetch_hdr().
940 */
941static int smp_fetch_chn_hdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
942{
943 kw = ((smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ ? "req.hdr" : "res.hdr");
944 return smp_fetch_hdr(args, smp, kw, private);
945}
946
Willy Tarreau79e57332018-10-02 16:01:16 +0200947/* 6. Check on HTTP header count. The number of occurrences is returned.
948 * Accepts exactly 1 argument of type string.
949 */
950static int smp_fetch_hdr_cnt(const struct arg *args, struct sample *smp, const char *kw, void *private)
951{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200952 /* possible keywords: req.hdr_cnt / hdr_cnt, res.hdr_cnt / shdr_cnt */
953 struct channel *chn = ((kw[0] == 'h' || kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +0200954 struct check *check = ((kw[0] == 's' || kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +0200955 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200956 struct http_hdr_ctx ctx;
957 struct ist name;
Willy Tarreau79e57332018-10-02 16:01:16 +0200958 int cnt;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200959
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200960 if (!htx)
961 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200962
Christopher Faulet623af932021-01-29 11:22:15 +0100963 if (args->type == ARGT_STR) {
Tim Duesterhus92c696e2021-02-28 16:11:36 +0100964 name = ist2(args->data.str.area, args->data.str.data);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200965 } else {
Tim Duesterhus68a088d2021-02-28 16:11:37 +0100966 name = IST_NULL;
Willy Tarreau79e57332018-10-02 16:01:16 +0200967 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200968
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200969 ctx.blk = NULL;
970 cnt = 0;
971 while (http_find_header(htx, name, &ctx, 0))
972 cnt++;
Willy Tarreau79e57332018-10-02 16:01:16 +0200973
974 smp->data.type = SMP_T_SINT;
975 smp->data.u.sint = cnt;
976 smp->flags = SMP_F_VOL_HDR;
977 return 1;
978}
979
980/* Fetch an HTTP header's integer value. The integer value is returned. It
981 * takes a mandatory argument of type string and an optional one of type int
982 * to designate a specific occurrence. It returns an unsigned integer, which
983 * may or may not be appropriate for everything.
984 */
985static int smp_fetch_hdr_val(const struct arg *args, struct sample *smp, const char *kw, void *private)
986{
987 int ret = smp_fetch_hdr(args, smp, kw, private);
988
989 if (ret > 0) {
990 smp->data.type = SMP_T_SINT;
991 smp->data.u.sint = strl2ic(smp->data.u.str.area,
992 smp->data.u.str.data);
993 }
994
995 return ret;
996}
997
998/* Fetch an HTTP header's IP value. takes a mandatory argument of type string
999 * and an optional one of type int to designate a specific occurrence.
Willy Tarreau7b0e00d2021-03-25 14:12:29 +01001000 * It returns an IPv4 or IPv6 address. Addresses surrounded by invalid chars
1001 * are rejected. However IPv4 addresses may be followed with a colon and a
1002 * valid port number.
Willy Tarreau79e57332018-10-02 16:01:16 +02001003 */
1004static int smp_fetch_hdr_ip(const struct arg *args, struct sample *smp, const char *kw, void *private)
1005{
Tim Duesterhus5cd00872020-06-26 15:44:48 +02001006 struct buffer *temp = get_trash_chunk();
Willy Tarreau7b0e00d2021-03-25 14:12:29 +01001007 int ret, len;
1008 int port;
Willy Tarreau79e57332018-10-02 16:01:16 +02001009
1010 while ((ret = smp_fetch_hdr(args, smp, kw, private)) > 0) {
Tim Duesterhus5cd00872020-06-26 15:44:48 +02001011 if (smp->data.u.str.data < temp->size - 1) {
1012 memcpy(temp->area, smp->data.u.str.area,
1013 smp->data.u.str.data);
1014 temp->area[smp->data.u.str.data] = '\0';
Willy Tarreau7b0e00d2021-03-25 14:12:29 +01001015 len = url2ipv4((char *) temp->area, &smp->data.u.ipv4);
Willy Tarreau645dc082021-03-31 11:41:36 +02001016 if (len > 0 && len == smp->data.u.str.data) {
Willy Tarreau7b0e00d2021-03-25 14:12:29 +01001017 /* plain IPv4 address */
1018 smp->data.type = SMP_T_IPV4;
1019 break;
1020 } else if (len > 0 && temp->area[len] == ':' &&
1021 strl2irc(temp->area + len + 1, smp->data.u.str.data - len - 1, &port) == 0 &&
1022 port >= 0 && port <= 65535) {
1023 /* IPv4 address suffixed with ':' followed by a valid port number */
Tim Duesterhus5cd00872020-06-26 15:44:48 +02001024 smp->data.type = SMP_T_IPV4;
1025 break;
1026 } else if (inet_pton(AF_INET6, temp->area, &smp->data.u.ipv6)) {
1027 smp->data.type = SMP_T_IPV6;
1028 break;
Willy Tarreau79e57332018-10-02 16:01:16 +02001029 }
1030 }
1031
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001032 /* if the header doesn't match an IP address, fetch next one */
1033 if (!(smp->flags & SMP_F_NOT_LAST))
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001034 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001035 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001036 return ret;
1037}
Willy Tarreau79e57332018-10-02 16:01:16 +02001038
Christopher Faulete720c322020-09-02 17:25:18 +02001039/* 8. Check on URI PATH. A pointer to the PATH is stored. The path starts at the
1040 * first '/' after the possible hostname. It ends before the possible '?' except
1041 * for 'pathq' keyword.
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001042 */
1043static int smp_fetch_path(const struct arg *args, struct sample *smp, const char *kw, void *private)
1044{
1045 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001046 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001047 struct htx_sl *sl;
1048 struct ist path;
Amaury Denoyellec453f952021-07-06 11:40:12 +02001049 struct http_uri_parser parser;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001050
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001051 if (!htx)
1052 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001053
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001054 sl = http_get_stline(htx);
Amaury Denoyellec453f952021-07-06 11:40:12 +02001055 parser = http_uri_parser_init(htx_sl_req_uri(sl));
Christopher Faulete720c322020-09-02 17:25:18 +02001056
Yves Lafonb4d37082021-02-11 11:01:28 +01001057 if (kw[4] == 'q' && (kw[0] == 'p' || kw[0] == 'b')) // pathq or baseq
Amaury Denoyellec453f952021-07-06 11:40:12 +02001058 path = http_parse_path(&parser);
Christopher Faulete720c322020-09-02 17:25:18 +02001059 else
Amaury Denoyellec453f952021-07-06 11:40:12 +02001060 path = iststop(http_parse_path(&parser), '?');
Christopher Faulete720c322020-09-02 17:25:18 +02001061
Tim Duesterhused526372020-03-05 17:56:33 +01001062 if (!isttest(path))
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001063 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001064
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001065 /* OK, we got the '/' ! */
1066 smp->data.type = SMP_T_STR;
1067 smp->data.u.str.area = path.ptr;
Jerome Magnin4fb196c2020-02-21 10:49:12 +01001068 smp->data.u.str.data = path.len;
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001069 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
Willy Tarreau79e57332018-10-02 16:01:16 +02001070 return 1;
1071}
1072
1073/* This produces a concatenation of the first occurrence of the Host header
1074 * followed by the path component if it begins with a slash ('/'). This means
1075 * that '*' will not be added, resulting in exactly the first Host entry.
1076 * If no Host header is found, then the path is returned as-is. The returned
1077 * value is stored in the trash so it does not need to be marked constant.
1078 * The returned sample is of type string.
1079 */
1080static int smp_fetch_base(const struct arg *args, struct sample *smp, const char *kw, void *private)
1081{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001082 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001083 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001084 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +02001085 struct buffer *temp;
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001086 struct http_hdr_ctx ctx;
1087 struct ist path;
Amaury Denoyellec453f952021-07-06 11:40:12 +02001088 struct http_uri_parser parser;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001089
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001090 if (!htx)
1091 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001092
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001093 ctx.blk = NULL;
1094 if (!http_find_header(htx, ist("Host"), &ctx, 0) || !ctx.value.len)
1095 return smp_fetch_path(args, smp, kw, private);
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001096
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001097 /* OK we have the header value in ctx.value */
1098 temp = get_trash_chunk();
Tim Duesterhus77508502022-03-15 13:11:06 +01001099 chunk_istcat(temp, ctx.value);
Willy Tarreau79e57332018-10-02 16:01:16 +02001100
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001101 /* now retrieve the path */
1102 sl = http_get_stline(htx);
Amaury Denoyellec453f952021-07-06 11:40:12 +02001103 parser = http_uri_parser_init(htx_sl_req_uri(sl));
1104 path = http_parse_path(&parser);
Tim Duesterhused526372020-03-05 17:56:33 +01001105 if (isttest(path)) {
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001106 size_t len;
Willy Tarreau79e57332018-10-02 16:01:16 +02001107
Yves Lafonb4d37082021-02-11 11:01:28 +01001108 if (kw[4] == 'q' && kw[0] == 'b') { // baseq
1109 len = path.len;
1110 } else {
1111 for (len = 0; len < path.len && *(path.ptr + len) != '?'; len++)
1112 ;
1113 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001114
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001115 if (len && *(path.ptr) == '/')
1116 chunk_memcat(temp, path.ptr, len);
Willy Tarreau79e57332018-10-02 16:01:16 +02001117 }
1118
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001119 smp->data.type = SMP_T_STR;
1120 smp->data.u.str = *temp;
Willy Tarreau79e57332018-10-02 16:01:16 +02001121 smp->flags = SMP_F_VOL_1ST;
1122 return 1;
1123}
1124
1125/* This produces a 32-bit hash of the concatenation of the first occurrence of
1126 * the Host header followed by the path component if it begins with a slash ('/').
1127 * This means that '*' will not be added, resulting in exactly the first Host
1128 * entry. If no Host header is found, then the path is used. The resulting value
1129 * is hashed using the path hash followed by a full avalanche hash and provides a
1130 * 32-bit integer value. This fetch is useful for tracking per-path activity on
1131 * high-traffic sites without having to store whole paths.
1132 */
1133static int smp_fetch_base32(const struct arg *args, struct sample *smp, const char *kw, void *private)
1134{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001135 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001136 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001137 struct htx_sl *sl;
1138 struct http_hdr_ctx ctx;
1139 struct ist path;
Willy Tarreau79e57332018-10-02 16:01:16 +02001140 unsigned int hash = 0;
Amaury Denoyellec453f952021-07-06 11:40:12 +02001141 struct http_uri_parser parser;
Willy Tarreau79e57332018-10-02 16:01:16 +02001142
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001143 if (!htx)
1144 return 0;
Dragan Dosen8861e1c2019-02-12 19:50:31 +01001145
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001146 ctx.blk = NULL;
1147 if (http_find_header(htx, ist("Host"), &ctx, 0)) {
1148 /* OK we have the header value in ctx.value */
1149 while (ctx.value.len--)
1150 hash = *(ctx.value.ptr++) + (hash << 6) + (hash << 16) - hash;
Willy Tarreau79e57332018-10-02 16:01:16 +02001151 }
1152
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001153 /* now retrieve the path */
1154 sl = http_get_stline(htx);
Amaury Denoyellec453f952021-07-06 11:40:12 +02001155 parser = http_uri_parser_init(htx_sl_req_uri(sl));
1156 path = http_parse_path(&parser);
Tim Duesterhused526372020-03-05 17:56:33 +01001157 if (isttest(path)) {
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001158 size_t len;
Willy Tarreau79e57332018-10-02 16:01:16 +02001159
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001160 for (len = 0; len < path.len && *(path.ptr + len) != '?'; len++)
1161 ;
Willy Tarreau79e57332018-10-02 16:01:16 +02001162
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001163 if (len && *(path.ptr) == '/') {
1164 while (len--)
1165 hash = *(path.ptr++) + (hash << 6) + (hash << 16) - hash;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001166 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001167 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001168
Willy Tarreau79e57332018-10-02 16:01:16 +02001169 hash = full_hash(hash);
1170
1171 smp->data.type = SMP_T_SINT;
1172 smp->data.u.sint = hash;
1173 smp->flags = SMP_F_VOL_1ST;
1174 return 1;
1175}
1176
1177/* This concatenates the source address with the 32-bit hash of the Host and
1178 * path as returned by smp_fetch_base32(). The idea is to have per-source and
1179 * per-path counters. The result is a binary block from 8 to 20 bytes depending
1180 * on the source address length. The path hash is stored before the address so
1181 * that in environments where IPv6 is insignificant, truncating the output to
1182 * 8 bytes would still work.
1183 */
1184static int smp_fetch_base32_src(const struct arg *args, struct sample *smp, const char *kw, void *private)
1185{
Willy Tarreaud68ff012022-05-27 08:57:21 +02001186 const struct sockaddr_storage *src = (smp->strm ? sc_src(smp->strm->scf) : NULL);
Willy Tarreau79e57332018-10-02 16:01:16 +02001187 struct buffer *temp;
Willy Tarreau79e57332018-10-02 16:01:16 +02001188
Christopher Faulet6fc817a2021-10-25 07:48:27 +02001189 if (!src)
Willy Tarreau79e57332018-10-02 16:01:16 +02001190 return 0;
1191
1192 if (!smp_fetch_base32(args, smp, kw, private))
1193 return 0;
1194
1195 temp = get_trash_chunk();
1196 *(unsigned int *) temp->area = htonl(smp->data.u.sint);
1197 temp->data += sizeof(unsigned int);
1198
Christopher Faulet6fc817a2021-10-25 07:48:27 +02001199 switch (src->ss_family) {
Willy Tarreau79e57332018-10-02 16:01:16 +02001200 case AF_INET:
1201 memcpy(temp->area + temp->data,
Christopher Faulet6fc817a2021-10-25 07:48:27 +02001202 &((struct sockaddr_in *)src)->sin_addr,
Willy Tarreau79e57332018-10-02 16:01:16 +02001203 4);
1204 temp->data += 4;
1205 break;
1206 case AF_INET6:
1207 memcpy(temp->area + temp->data,
Christopher Faulet6fc817a2021-10-25 07:48:27 +02001208 &((struct sockaddr_in6 *)src)->sin6_addr,
Willy Tarreau79e57332018-10-02 16:01:16 +02001209 16);
1210 temp->data += 16;
1211 break;
1212 default:
1213 return 0;
1214 }
1215
1216 smp->data.u.str = *temp;
1217 smp->data.type = SMP_T_BIN;
1218 return 1;
1219}
1220
1221/* Extracts the query string, which comes after the question mark '?'. If no
1222 * question mark is found, nothing is returned. Otherwise it returns a sample
1223 * of type string carrying the whole query string.
1224 */
1225static int smp_fetch_query(const struct arg *args, struct sample *smp, const char *kw, void *private)
1226{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001227 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001228 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001229 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +02001230 char *ptr, *end;
1231
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001232 if (!htx)
1233 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001234
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001235 sl = http_get_stline(htx);
1236 ptr = HTX_SL_REQ_UPTR(sl);
1237 end = HTX_SL_REQ_UPTR(sl) + HTX_SL_REQ_ULEN(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +02001238
1239 /* look up the '?' */
1240 do {
1241 if (ptr == end)
1242 return 0;
1243 } while (*ptr++ != '?');
1244
1245 smp->data.type = SMP_T_STR;
1246 smp->data.u.str.area = ptr;
1247 smp->data.u.str.data = end - ptr;
1248 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
1249 return 1;
1250}
1251
1252static int smp_fetch_proto_http(const struct arg *args, struct sample *smp, const char *kw, void *private)
1253{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001254 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001255 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 0);
Willy Tarreau79e57332018-10-02 16:01:16 +02001256
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001257 if (!htx)
1258 return 0;
1259 smp->data.type = SMP_T_BOOL;
Willy Tarreau79e57332018-10-02 16:01:16 +02001260 smp->data.u.sint = 1;
1261 return 1;
1262}
1263
1264/* return a valid test if the current request is the first one on the connection */
1265static int smp_fetch_http_first_req(const struct arg *args, struct sample *smp, const char *kw, void *private)
1266{
Willy Tarreau79512b62020-04-29 11:52:13 +02001267 if (!smp->strm)
1268 return 0;
1269
Willy Tarreau79e57332018-10-02 16:01:16 +02001270 smp->data.type = SMP_T_BOOL;
1271 smp->data.u.sint = !(smp->strm->txn->flags & TX_NOT_FIRST);
1272 return 1;
1273}
1274
Christopher Fauleta4063562019-08-02 11:51:37 +02001275/* Fetch the authentication method if there is an Authorization header. It
1276 * relies on get_http_auth()
1277 */
1278static int smp_fetch_http_auth_type(const struct arg *args, struct sample *smp, const char *kw, void *private)
1279{
1280 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001281 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Fauleta4063562019-08-02 11:51:37 +02001282 struct http_txn *txn;
1283
1284 if (!htx)
1285 return 0;
1286
1287 txn = smp->strm->txn;
1288 if (!get_http_auth(smp, htx))
1289 return 0;
1290
1291 switch (txn->auth.method) {
1292 case HTTP_AUTH_BASIC:
1293 smp->data.u.str.area = "Basic";
1294 smp->data.u.str.data = 5;
1295 break;
1296 case HTTP_AUTH_DIGEST:
1297 /* Unexpected because not supported */
1298 smp->data.u.str.area = "Digest";
1299 smp->data.u.str.data = 6;
1300 break;
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +02001301 case HTTP_AUTH_BEARER:
1302 smp->data.u.str.area = "Bearer";
1303 smp->data.u.str.data = 6;
1304 break;
Christopher Fauleta4063562019-08-02 11:51:37 +02001305 default:
1306 return 0;
1307 }
1308
1309 smp->data.type = SMP_T_STR;
1310 smp->flags = SMP_F_CONST;
1311 return 1;
1312}
1313
1314/* Fetch the user supplied if there is an Authorization header. It relies on
1315 * get_http_auth()
1316 */
1317static int smp_fetch_http_auth_user(const struct arg *args, struct sample *smp, const char *kw, void *private)
1318{
1319 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001320 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Fauleta4063562019-08-02 11:51:37 +02001321 struct http_txn *txn;
1322
1323 if (!htx)
1324 return 0;
1325
1326 txn = smp->strm->txn;
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +02001327 if (!get_http_auth(smp, htx) || txn->auth.method != HTTP_AUTH_BASIC)
Christopher Fauleta4063562019-08-02 11:51:37 +02001328 return 0;
1329
1330 smp->data.type = SMP_T_STR;
1331 smp->data.u.str.area = txn->auth.user;
1332 smp->data.u.str.data = strlen(txn->auth.user);
1333 smp->flags = SMP_F_CONST;
1334 return 1;
1335}
1336
1337/* Fetch the password supplied if there is an Authorization header. It relies on
1338 * get_http_auth()
1339 */
1340static int smp_fetch_http_auth_pass(const struct arg *args, struct sample *smp, const char *kw, void *private)
1341{
1342 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001343 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Fauleta4063562019-08-02 11:51:37 +02001344 struct http_txn *txn;
1345
1346 if (!htx)
1347 return 0;
1348
1349 txn = smp->strm->txn;
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +02001350 if (!get_http_auth(smp, htx) || txn->auth.method != HTTP_AUTH_BASIC)
Christopher Fauleta4063562019-08-02 11:51:37 +02001351 return 0;
1352
1353 smp->data.type = SMP_T_STR;
1354 smp->data.u.str.area = txn->auth.pass;
1355 smp->data.u.str.data = strlen(txn->auth.pass);
1356 smp->flags = SMP_F_CONST;
1357 return 1;
1358}
1359
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +02001360static int smp_fetch_http_auth_bearer(const struct arg *args, struct sample *smp, const char *kw, void *private)
1361{
1362 struct channel *chn = SMP_REQ_CHN(smp);
1363 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
1364 struct http_txn *txn;
1365 struct buffer bearer_val = {};
1366
1367 if (!htx)
1368 return 0;
1369
1370 if (args->type == ARGT_STR) {
1371 struct http_hdr_ctx ctx;
1372 struct ist hdr_name = ist2(args->data.str.area, args->data.str.data);
1373
1374 ctx.blk = NULL;
1375 if (http_find_header(htx, hdr_name, &ctx, 0)) {
Remi Tricot-Le Breton7da35bf2021-10-29 15:25:19 +02001376 struct ist type = istsplit(&ctx.value, ' ');
1377
1378 /* There must be "at least" one space character between
1379 * the scheme and the following value so ctx.value might
1380 * still have leading spaces here (see RFC7235).
1381 */
1382 ctx.value = istskip(ctx.value, ' ');
1383
1384 if (isteqi(type, ist("Bearer")) && istlen(ctx.value))
1385 chunk_initlen(&bearer_val, istptr(ctx.value), 0, istlen(ctx.value));
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +02001386 }
1387 }
1388 else {
1389 txn = smp->strm->txn;
1390 if (!get_http_auth(smp, htx) || txn->auth.method != HTTP_AUTH_BEARER)
1391 return 0;
1392
1393 bearer_val = txn->auth.method_data;
1394 }
1395
1396 smp->data.type = SMP_T_STR;
1397 smp->data.u.str = bearer_val;
1398 smp->flags = SMP_F_CONST;
1399 return 1;
1400}
1401
Willy Tarreau79e57332018-10-02 16:01:16 +02001402/* Accepts exactly 1 argument of type userlist */
1403static int smp_fetch_http_auth(const struct arg *args, struct sample *smp, const char *kw, void *private)
1404{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001405 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001406 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Willy Tarreau79e57332018-10-02 16:01:16 +02001407
Christopher Faulet623af932021-01-29 11:22:15 +01001408 if (args->type != ARGT_USR)
Willy Tarreau79e57332018-10-02 16:01:16 +02001409 return 0;
1410
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001411 if (!htx)
1412 return 0;
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +02001413 if (!get_http_auth(smp, htx) || smp->strm->txn->auth.method != HTTP_AUTH_BASIC)
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001414 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001415
1416 smp->data.type = SMP_T_BOOL;
1417 smp->data.u.sint = check_user(args->data.usr, smp->strm->txn->auth.user,
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001418 smp->strm->txn->auth.pass);
Willy Tarreau79e57332018-10-02 16:01:16 +02001419 return 1;
1420}
1421
1422/* Accepts exactly 1 argument of type userlist */
1423static int smp_fetch_http_auth_grp(const struct arg *args, struct sample *smp, const char *kw, void *private)
1424{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001425 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001426 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet89dc4992019-04-17 12:02:59 +02001427
Christopher Faulet623af932021-01-29 11:22:15 +01001428 if (args->type != ARGT_USR)
Willy Tarreau79e57332018-10-02 16:01:16 +02001429 return 0;
1430
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001431 if (!htx)
1432 return 0;
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +02001433 if (!get_http_auth(smp, htx) || smp->strm->txn->auth.method != HTTP_AUTH_BASIC)
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001434 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001435
Willy Tarreau79e57332018-10-02 16:01:16 +02001436 /* if the user does not belong to the userlist or has a wrong password,
1437 * report that it unconditionally does not match. Otherwise we return
1438 * a string containing the username.
1439 */
1440 if (!check_user(args->data.usr, smp->strm->txn->auth.user,
1441 smp->strm->txn->auth.pass))
1442 return 0;
1443
1444 /* pat_match_auth() will need the user list */
1445 smp->ctx.a[0] = args->data.usr;
1446
1447 smp->data.type = SMP_T_STR;
1448 smp->flags = SMP_F_CONST;
1449 smp->data.u.str.area = smp->strm->txn->auth.user;
1450 smp->data.u.str.data = strlen(smp->strm->txn->auth.user);
1451
1452 return 1;
1453}
1454
1455/* Fetch a captured HTTP request header. The index is the position of
1456 * the "capture" option in the configuration file
1457 */
1458static int smp_fetch_capture_req_hdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
1459{
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001460 struct proxy *fe;
Willy Tarreau79e57332018-10-02 16:01:16 +02001461 int idx;
1462
Christopher Faulet623af932021-01-29 11:22:15 +01001463 if (args->type != ARGT_SINT)
Willy Tarreau79e57332018-10-02 16:01:16 +02001464 return 0;
1465
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001466 if (!smp->strm)
1467 return 0;
1468
1469 fe = strm_fe(smp->strm);
Willy Tarreau79e57332018-10-02 16:01:16 +02001470 idx = args->data.sint;
1471
1472 if (idx > (fe->nb_req_cap - 1) || smp->strm->req_cap == NULL || smp->strm->req_cap[idx] == NULL)
1473 return 0;
1474
1475 smp->data.type = SMP_T_STR;
1476 smp->flags |= SMP_F_CONST;
1477 smp->data.u.str.area = smp->strm->req_cap[idx];
1478 smp->data.u.str.data = strlen(smp->strm->req_cap[idx]);
1479
1480 return 1;
1481}
1482
1483/* Fetch a captured HTTP response header. The index is the position of
1484 * the "capture" option in the configuration file
1485 */
1486static int smp_fetch_capture_res_hdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
1487{
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001488 struct proxy *fe;
Willy Tarreau79e57332018-10-02 16:01:16 +02001489 int idx;
1490
Christopher Faulet623af932021-01-29 11:22:15 +01001491 if (args->type != ARGT_SINT)
Willy Tarreau79e57332018-10-02 16:01:16 +02001492 return 0;
1493
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001494 if (!smp->strm)
1495 return 0;
1496
1497 fe = strm_fe(smp->strm);
Willy Tarreau79e57332018-10-02 16:01:16 +02001498 idx = args->data.sint;
1499
1500 if (idx > (fe->nb_rsp_cap - 1) || smp->strm->res_cap == NULL || smp->strm->res_cap[idx] == NULL)
1501 return 0;
1502
1503 smp->data.type = SMP_T_STR;
1504 smp->flags |= SMP_F_CONST;
1505 smp->data.u.str.area = smp->strm->res_cap[idx];
1506 smp->data.u.str.data = strlen(smp->strm->res_cap[idx]);
1507
1508 return 1;
1509}
1510
1511/* Extracts the METHOD in the HTTP request, the txn->uri should be filled before the call */
1512static int smp_fetch_capture_req_method(const struct arg *args, struct sample *smp, const char *kw, void *private)
1513{
1514 struct buffer *temp;
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001515 struct http_txn *txn;
Willy Tarreau79e57332018-10-02 16:01:16 +02001516 char *ptr;
1517
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001518 if (!smp->strm)
1519 return 0;
1520
1521 txn = smp->strm->txn;
Willy Tarreau79e57332018-10-02 16:01:16 +02001522 if (!txn || !txn->uri)
1523 return 0;
1524
1525 ptr = txn->uri;
1526
1527 while (*ptr != ' ' && *ptr != '\0') /* find first space */
1528 ptr++;
1529
1530 temp = get_trash_chunk();
1531 temp->area = txn->uri;
1532 temp->data = ptr - txn->uri;
1533 smp->data.u.str = *temp;
1534 smp->data.type = SMP_T_STR;
1535 smp->flags = SMP_F_CONST;
1536
1537 return 1;
1538
1539}
1540
1541/* Extracts the path in the HTTP request, the txn->uri should be filled before the call */
1542static int smp_fetch_capture_req_uri(const struct arg *args, struct sample *smp, const char *kw, void *private)
1543{
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001544 struct http_txn *txn;
Willy Tarreau79e57332018-10-02 16:01:16 +02001545 struct ist path;
1546 const char *ptr;
Amaury Denoyellec453f952021-07-06 11:40:12 +02001547 struct http_uri_parser parser;
Willy Tarreau79e57332018-10-02 16:01:16 +02001548
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001549 if (!smp->strm)
1550 return 0;
1551
1552 txn = smp->strm->txn;
Willy Tarreau79e57332018-10-02 16:01:16 +02001553 if (!txn || !txn->uri)
1554 return 0;
1555
1556 ptr = txn->uri;
1557
1558 while (*ptr != ' ' && *ptr != '\0') /* find first space */
1559 ptr++;
1560
1561 if (!*ptr)
1562 return 0;
1563
Christopher Faulet78337bb2018-11-15 14:35:18 +01001564 /* skip the first space and find space after URI */
1565 path = ist2(++ptr, 0);
1566 while (*ptr != ' ' && *ptr != '\0')
1567 ptr++;
1568 path.len = ptr - path.ptr;
Willy Tarreau79e57332018-10-02 16:01:16 +02001569
Amaury Denoyellec453f952021-07-06 11:40:12 +02001570 parser = http_uri_parser_init(path);
1571 path = http_parse_path(&parser);
Tim Duesterhused526372020-03-05 17:56:33 +01001572 if (!isttest(path))
Willy Tarreau79e57332018-10-02 16:01:16 +02001573 return 0;
1574
1575 smp->data.u.str.area = path.ptr;
1576 smp->data.u.str.data = path.len;
1577 smp->data.type = SMP_T_STR;
1578 smp->flags = SMP_F_CONST;
1579
1580 return 1;
1581}
1582
1583/* Retrieves the HTTP version from the request (either 1.0 or 1.1) and emits it
1584 * as a string (either "HTTP/1.0" or "HTTP/1.1").
1585 */
1586static int smp_fetch_capture_req_ver(const struct arg *args, struct sample *smp, const char *kw, void *private)
1587{
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001588 struct http_txn *txn;
1589
1590 if (!smp->strm)
1591 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001592
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001593 txn = smp->strm->txn;
Christopher Faulet09f88362021-04-01 16:00:29 +02001594 if (!txn || txn->req.msg_state < HTTP_MSG_BODY)
Willy Tarreau79e57332018-10-02 16:01:16 +02001595 return 0;
1596
1597 if (txn->req.flags & HTTP_MSGF_VER_11)
1598 smp->data.u.str.area = "HTTP/1.1";
1599 else
1600 smp->data.u.str.area = "HTTP/1.0";
1601
1602 smp->data.u.str.data = 8;
1603 smp->data.type = SMP_T_STR;
1604 smp->flags = SMP_F_CONST;
1605 return 1;
1606
1607}
1608
1609/* Retrieves the HTTP version from the response (either 1.0 or 1.1) and emits it
1610 * as a string (either "HTTP/1.0" or "HTTP/1.1").
1611 */
1612static int smp_fetch_capture_res_ver(const struct arg *args, struct sample *smp, const char *kw, void *private)
1613{
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001614 struct http_txn *txn;
1615
1616 if (!smp->strm)
1617 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001618
Willy Tarreau0898c2d2020-04-29 11:44:54 +02001619 txn = smp->strm->txn;
Christopher Faulet09f88362021-04-01 16:00:29 +02001620 if (!txn || txn->rsp.msg_state < HTTP_MSG_BODY)
Willy Tarreau79e57332018-10-02 16:01:16 +02001621 return 0;
1622
1623 if (txn->rsp.flags & HTTP_MSGF_VER_11)
1624 smp->data.u.str.area = "HTTP/1.1";
1625 else
1626 smp->data.u.str.area = "HTTP/1.0";
1627
1628 smp->data.u.str.data = 8;
1629 smp->data.type = SMP_T_STR;
1630 smp->flags = SMP_F_CONST;
1631 return 1;
1632
1633}
1634
1635/* Iterate over all cookies present in a message. The context is stored in
1636 * smp->ctx.a[0] for the in-header position, smp->ctx.a[1] for the
1637 * end-of-header-value, and smp->ctx.a[2] for the hdr_ctx. Depending on
1638 * the direction, multiple cookies may be parsed on the same line or not.
Maciej Zdebdea7c202020-11-13 09:38:06 +00001639 * If provided, the searched cookie name is in args, in args->data.str. If
1640 * the input options indicate that no iterating is desired, then only last
1641 * value is fetched if any. If no cookie name is provided, the first cookie
1642 * value found is fetched. The returned sample is of type CSTR. Can be used
1643 * to parse cookies in other files.
Willy Tarreau79e57332018-10-02 16:01:16 +02001644 */
1645static int smp_fetch_cookie(const struct arg *args, struct sample *smp, const char *kw, void *private)
1646{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001647 /* possible keywords: req.cookie / cookie / cook, res.cookie / scook / set-cookie */
1648 struct channel *chn = ((kw[0] == 'c' || kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +02001649 struct check *check = ((kw[0] == 's' || kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +02001650 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001651 struct http_hdr_ctx *ctx = smp->ctx.a[2];
1652 struct ist hdr;
Christopher Faulet97fc8da2020-11-13 13:41:04 +01001653 char *cook = NULL;
1654 size_t cook_l = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001655 int found = 0;
1656
Christopher Faulet623af932021-01-29 11:22:15 +01001657 if (args->type == ARGT_STR) {
Christopher Faulet97fc8da2020-11-13 13:41:04 +01001658 cook = args->data.str.area;
1659 cook_l = args->data.str.data;
1660 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001661
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001662 if (!ctx) {
1663 /* first call */
1664 ctx = &static_http_hdr_ctx;
1665 ctx->blk = NULL;
1666 smp->ctx.a[2] = ctx;
1667 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001668
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001669 if (!htx)
1670 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001671
Christopher Faulet16032ab2020-04-30 11:30:00 +02001672 hdr = (!(check || (chn && chn->flags & CF_ISRESP)) ? ist("Cookie") : ist("Set-Cookie"));
Willy Tarreau79e57332018-10-02 16:01:16 +02001673
Maciej Zdebdea7c202020-11-13 09:38:06 +00001674 /* OK so basically here, either we want only one value or we want to
1675 * iterate over all of them and we fetch the next one. In this last case
1676 * SMP_OPT_ITERATE option is set.
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001677 */
Willy Tarreau79e57332018-10-02 16:01:16 +02001678
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001679 if (!(smp->flags & SMP_F_NOT_LAST)) {
1680 /* search for the header from the beginning, we must first initialize
1681 * the search parameters.
Willy Tarreau79e57332018-10-02 16:01:16 +02001682 */
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001683 smp->ctx.a[0] = NULL;
1684 ctx->blk = NULL;
Willy Tarreau79e57332018-10-02 16:01:16 +02001685 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001686
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001687 smp->flags |= SMP_F_VOL_HDR;
1688 while (1) {
1689 /* Note: smp->ctx.a[0] == NULL every time we need to fetch a new header */
1690 if (!smp->ctx.a[0]) {
1691 if (!http_find_header(htx, hdr, ctx, 0))
1692 goto out;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001693
Christopher Faulet97fc8da2020-11-13 13:41:04 +01001694 if (ctx->value.len < cook_l + 1)
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001695 continue;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001696
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001697 smp->ctx.a[0] = ctx->value.ptr;
1698 smp->ctx.a[1] = smp->ctx.a[0] + ctx->value.len;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001699 }
1700
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001701 smp->data.type = SMP_T_STR;
1702 smp->flags |= SMP_F_CONST;
1703 smp->ctx.a[0] = http_extract_cookie_value(smp->ctx.a[0], smp->ctx.a[1],
Christopher Faulet97fc8da2020-11-13 13:41:04 +01001704 cook, cook_l,
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001705 (smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ,
1706 &smp->data.u.str.area,
1707 &smp->data.u.str.data);
1708 if (smp->ctx.a[0]) {
1709 found = 1;
Maciej Zdebdea7c202020-11-13 09:38:06 +00001710 if (smp->opt & SMP_OPT_ITERATE) {
1711 /* iterate on cookie value */
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001712 smp->flags |= SMP_F_NOT_LAST;
1713 return 1;
Willy Tarreau79e57332018-10-02 16:01:16 +02001714 }
Maciej Zdebdea7c202020-11-13 09:38:06 +00001715 if (args->data.str.data == 0) {
1716 /* No cookie name, first occurrence returned */
1717 break;
1718 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001719 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001720 /* if we're looking for last occurrence, let's loop */
Willy Tarreau79e57332018-10-02 16:01:16 +02001721 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001722
Willy Tarreau79e57332018-10-02 16:01:16 +02001723 /* all cookie headers and values were scanned. If we're looking for the
1724 * last occurrence, we may return it now.
1725 */
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001726 out:
Willy Tarreau79e57332018-10-02 16:01:16 +02001727 smp->flags &= ~SMP_F_NOT_LAST;
1728 return found;
1729}
1730
Christopher Fauletc1f40dd2019-05-16 10:07:30 +02001731/* Same than smp_fetch_cookie() but only relies on the sample direction to
1732 * choose the right channel. So instead of duplicating the code, we just change
1733 * the keyword and then fallback on smp_fetch_cookie().
1734 */
1735static int smp_fetch_chn_cookie(const struct arg *args, struct sample *smp, const char *kw, void *private)
1736{
1737 kw = ((smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ ? "req.cook" : "res.cook");
1738 return smp_fetch_cookie(args, smp, kw, private);
1739}
1740
Willy Tarreau79e57332018-10-02 16:01:16 +02001741/* Iterate over all cookies present in a request to count how many occurrences
1742 * match the name in args and args->data.str.len. If <multi> is non-null, then
1743 * multiple cookies may be parsed on the same line. The returned sample is of
1744 * type UINT. Accepts exactly 1 argument of type string.
1745 */
1746static int smp_fetch_cookie_cnt(const struct arg *args, struct sample *smp, const char *kw, void *private)
1747{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001748 /* possible keywords: req.cook_cnt / cook_cnt, res.cook_cnt / scook_cnt */
1749 struct channel *chn = ((kw[0] == 'c' || kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Fauletf98e6262020-05-06 09:42:04 +02001750 struct check *check = ((kw[0] == 's' || kw[2] == 's') ? objt_check(smp->sess->origin) : NULL);
Christopher Faulet16032ab2020-04-30 11:30:00 +02001751 struct htx *htx = smp_prefetch_htx(smp, chn, check, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001752 struct http_hdr_ctx ctx;
1753 struct ist hdr;
Willy Tarreau79e57332018-10-02 16:01:16 +02001754 char *val_beg, *val_end;
Christopher Faulet97fc8da2020-11-13 13:41:04 +01001755 char *cook = NULL;
1756 size_t cook_l = 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001757 int cnt;
Willy Tarreau79e57332018-10-02 16:01:16 +02001758
Christopher Faulet623af932021-01-29 11:22:15 +01001759 if (args->type == ARGT_STR){
Christopher Faulet97fc8da2020-11-13 13:41:04 +01001760 cook = args->data.str.area;
1761 cook_l = args->data.str.data;
1762 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001763
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001764 if (!htx)
1765 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001766
Christopher Faulet16032ab2020-04-30 11:30:00 +02001767 hdr = (!(check || (chn && chn->flags & CF_ISRESP)) ? ist("Cookie") : ist("Set-Cookie"));
Willy Tarreau79e57332018-10-02 16:01:16 +02001768
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001769 val_end = val_beg = NULL;
1770 ctx.blk = NULL;
1771 cnt = 0;
1772 while (1) {
1773 /* Note: val_beg == NULL every time we need to fetch a new header */
1774 if (!val_beg) {
1775 if (!http_find_header(htx, hdr, &ctx, 0))
1776 break;
Willy Tarreau79e57332018-10-02 16:01:16 +02001777
Christopher Faulet97fc8da2020-11-13 13:41:04 +01001778 if (ctx.value.len < cook_l + 1)
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001779 continue;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001780
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001781 val_beg = ctx.value.ptr;
1782 val_end = val_beg + ctx.value.len;
Willy Tarreau79e57332018-10-02 16:01:16 +02001783 }
1784
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001785 smp->data.type = SMP_T_STR;
1786 smp->flags |= SMP_F_CONST;
1787 while ((val_beg = http_extract_cookie_value(val_beg, val_end,
Christopher Faulet97fc8da2020-11-13 13:41:04 +01001788 cook, cook_l,
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001789 (smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ,
1790 &smp->data.u.str.area,
1791 &smp->data.u.str.data))) {
1792 cnt++;
Willy Tarreau79e57332018-10-02 16:01:16 +02001793 }
1794 }
1795
1796 smp->data.type = SMP_T_SINT;
1797 smp->data.u.sint = cnt;
1798 smp->flags |= SMP_F_VOL_HDR;
1799 return 1;
1800}
1801
1802/* Fetch an cookie's integer value. The integer value is returned. It
1803 * takes a mandatory argument of type string. It relies on smp_fetch_cookie().
1804 */
1805static int smp_fetch_cookie_val(const struct arg *args, struct sample *smp, const char *kw, void *private)
1806{
1807 int ret = smp_fetch_cookie(args, smp, kw, private);
1808
1809 if (ret > 0) {
1810 smp->data.type = SMP_T_SINT;
1811 smp->data.u.sint = strl2ic(smp->data.u.str.area,
1812 smp->data.u.str.data);
1813 }
1814
1815 return ret;
1816}
1817
1818/************************************************************************/
1819/* The code below is dedicated to sample fetches */
1820/************************************************************************/
1821
1822/* This scans a URL-encoded query string. It takes an optionally wrapping
Ilya Shipitsin46a030c2020-07-05 16:36:08 +05001823 * string whose first contiguous chunk has its beginning in ctx->a[0] and end
Willy Tarreau79e57332018-10-02 16:01:16 +02001824 * in ctx->a[1], and the optional second part in (ctx->a[2]..ctx->a[3]). The
1825 * pointers are updated for next iteration before leaving.
1826 */
1827static int smp_fetch_param(char delim, const char *name, int name_len, const struct arg *args, struct sample *smp, const char *kw, void *private)
1828{
1829 const char *vstart, *vend;
1830 struct buffer *temp;
1831 const char **chunks = (const char **)smp->ctx.a;
1832
1833 if (!http_find_next_url_param(chunks, name, name_len,
1834 &vstart, &vend, delim))
1835 return 0;
1836
1837 /* Create sample. If the value is contiguous, return the pointer as CONST,
1838 * if the value is wrapped, copy-it in a buffer.
1839 */
1840 smp->data.type = SMP_T_STR;
1841 if (chunks[2] &&
1842 vstart >= chunks[0] && vstart <= chunks[1] &&
1843 vend >= chunks[2] && vend <= chunks[3]) {
1844 /* Wrapped case. */
1845 temp = get_trash_chunk();
1846 memcpy(temp->area, vstart, chunks[1] - vstart);
1847 memcpy(temp->area + ( chunks[1] - vstart ), chunks[2],
1848 vend - chunks[2]);
1849 smp->data.u.str.area = temp->area;
1850 smp->data.u.str.data = ( chunks[1] - vstart ) + ( vend - chunks[2] );
1851 } else {
1852 /* Contiguous case. */
1853 smp->data.u.str.area = (char *)vstart;
1854 smp->data.u.str.data = vend - vstart;
1855 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
1856 }
1857
1858 /* Update context, check wrapping. */
1859 chunks[0] = vend;
1860 if (chunks[2] && vend >= chunks[2] && vend <= chunks[3]) {
1861 chunks[1] = chunks[3];
1862 chunks[2] = NULL;
1863 }
1864
1865 if (chunks[0] < chunks[1])
1866 smp->flags |= SMP_F_NOT_LAST;
1867
1868 return 1;
1869}
1870
1871/* This function iterates over each parameter of the query string. It uses
1872 * ctx->a[0] and ctx->a[1] to store the beginning and end of the current
1873 * parameter. Since it uses smp_fetch_param(), ctx->a[2..3] are both NULL.
1874 * An optional parameter name is passed in args[0], otherwise any parameter is
1875 * considered. It supports an optional delimiter argument for the beginning of
1876 * the string in args[1], which defaults to "?".
1877 */
1878static int smp_fetch_url_param(const struct arg *args, struct sample *smp, const char *kw, void *private)
1879{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001880 struct channel *chn = SMP_REQ_CHN(smp);
Willy Tarreau79e57332018-10-02 16:01:16 +02001881 char delim = '?';
1882 const char *name;
1883 int name_len;
1884
Christopher Faulet623af932021-01-29 11:22:15 +01001885 if ((args[0].type && args[0].type != ARGT_STR) ||
Willy Tarreau79e57332018-10-02 16:01:16 +02001886 (args[1].type && args[1].type != ARGT_STR))
1887 return 0;
1888
1889 name = "";
1890 name_len = 0;
1891 if (args->type == ARGT_STR) {
1892 name = args->data.str.area;
1893 name_len = args->data.str.data;
1894 }
1895
1896 if (args[1].type)
1897 delim = *args[1].data.str.area;
1898
1899 if (!smp->ctx.a[0]) { // first call, find the query string
Christopher Faulet778f5ed2020-04-29 15:51:55 +02001900 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001901 struct htx_sl *sl;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001902
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001903 if (!htx)
1904 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001905
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001906 sl = http_get_stline(htx);
1907 smp->ctx.a[0] = http_find_param_list(HTX_SL_REQ_UPTR(sl), HTX_SL_REQ_ULEN(sl), delim);
1908 if (!smp->ctx.a[0])
1909 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001910
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001911 smp->ctx.a[1] = HTX_SL_REQ_UPTR(sl) + HTX_SL_REQ_ULEN(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +02001912
1913 /* Assume that the context is filled with NULL pointer
1914 * before the first call.
1915 * smp->ctx.a[2] = NULL;
1916 * smp->ctx.a[3] = NULL;
1917 */
1918 }
1919
1920 return smp_fetch_param(delim, name, name_len, args, smp, kw, private);
1921}
1922
1923/* This function iterates over each parameter of the body. This requires
1924 * that the body has been waited for using http-buffer-request. It uses
1925 * ctx->a[0] and ctx->a[1] to store the beginning and end of the first
Ilya Shipitsin46a030c2020-07-05 16:36:08 +05001926 * contiguous part of the body, and optionally ctx->a[2..3] to reference the
Willy Tarreau79e57332018-10-02 16:01:16 +02001927 * optional second part if the body wraps at the end of the buffer. An optional
1928 * parameter name is passed in args[0], otherwise any parameter is considered.
1929 */
1930static int smp_fetch_body_param(const struct arg *args, struct sample *smp, const char *kw, void *private)
1931{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001932 struct channel *chn = SMP_REQ_CHN(smp);
Willy Tarreau79e57332018-10-02 16:01:16 +02001933 const char *name;
1934 int name_len;
1935
Christopher Faulet623af932021-01-29 11:22:15 +01001936 if (args[0].type && args[0].type != ARGT_STR)
Willy Tarreau79e57332018-10-02 16:01:16 +02001937 return 0;
1938
1939 name = "";
1940 name_len = 0;
1941 if (args[0].type == ARGT_STR) {
1942 name = args[0].data.str.area;
1943 name_len = args[0].data.str.data;
1944 }
1945
1946 if (!smp->ctx.a[0]) { // first call, find the query string
Christopher Faulete596d182020-05-05 17:46:34 +02001947 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001948 struct buffer *temp;
1949 int32_t pos;
Willy Tarreau79e57332018-10-02 16:01:16 +02001950
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001951 if (!htx)
1952 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001953
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001954 temp = get_trash_chunk();
1955 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
1956 struct htx_blk *blk = htx_get_blk(htx, pos);
1957 enum htx_blk_type type = htx_get_blk_type(blk);
Willy Tarreau79e57332018-10-02 16:01:16 +02001958
Christopher Fauletd1ac2b92020-12-02 19:12:22 +01001959 if (type == HTX_BLK_TLR || type == HTX_BLK_EOT)
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001960 break;
1961 if (type == HTX_BLK_DATA) {
Christopher Faulet53a899b2019-10-08 16:38:42 +02001962 if (!h1_format_htx_data(htx_get_blk_value(htx, blk), temp, 0))
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001963 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001964 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001965 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001966
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001967 smp->ctx.a[0] = temp->area;
1968 smp->ctx.a[1] = temp->area + temp->data;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001969
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001970 /* Assume that the context is filled with NULL pointer
1971 * before the first call.
1972 * smp->ctx.a[2] = NULL;
1973 * smp->ctx.a[3] = NULL;
1974 */
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001975
Willy Tarreau79e57332018-10-02 16:01:16 +02001976 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001977
Willy Tarreau79e57332018-10-02 16:01:16 +02001978 return smp_fetch_param('&', name, name_len, args, smp, kw, private);
1979}
1980
1981/* Return the signed integer value for the specified url parameter (see url_param
1982 * above).
1983 */
1984static int smp_fetch_url_param_val(const struct arg *args, struct sample *smp, const char *kw, void *private)
1985{
1986 int ret = smp_fetch_url_param(args, smp, kw, private);
1987
1988 if (ret > 0) {
1989 smp->data.type = SMP_T_SINT;
1990 smp->data.u.sint = strl2ic(smp->data.u.str.area,
1991 smp->data.u.str.data);
1992 }
1993
1994 return ret;
1995}
1996
1997/* This produces a 32-bit hash of the concatenation of the first occurrence of
1998 * the Host header followed by the path component if it begins with a slash ('/').
1999 * This means that '*' will not be added, resulting in exactly the first Host
2000 * entry. If no Host header is found, then the path is used. The resulting value
2001 * is hashed using the url hash followed by a full avalanche hash and provides a
2002 * 32-bit integer value. This fetch is useful for tracking per-URL activity on
2003 * high-traffic sites without having to store whole paths.
2004 * this differs from the base32 functions in that it includes the url parameters
2005 * as well as the path
2006 */
2007static int smp_fetch_url32(const struct arg *args, struct sample *smp, const char *kw, void *private)
2008{
Christopher Faulet89dc4992019-04-17 12:02:59 +02002009 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet778f5ed2020-04-29 15:51:55 +02002010 struct htx *htx = smp_prefetch_htx(smp, chn, NULL, 1);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02002011 struct http_hdr_ctx ctx;
2012 struct htx_sl *sl;
2013 struct ist path;
Willy Tarreau79e57332018-10-02 16:01:16 +02002014 unsigned int hash = 0;
Amaury Denoyellec453f952021-07-06 11:40:12 +02002015 struct http_uri_parser parser;
Willy Tarreau79e57332018-10-02 16:01:16 +02002016
Christopher Faulet6d1dd462019-07-15 14:36:03 +02002017 if (!htx)
2018 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02002019
Christopher Faulet6d1dd462019-07-15 14:36:03 +02002020 ctx.blk = NULL;
2021 if (http_find_header(htx, ist("Host"), &ctx, 1)) {
2022 /* OK we have the header value in ctx.value */
2023 while (ctx.value.len--)
2024 hash = *(ctx.value.ptr++) + (hash << 6) + (hash << 16) - hash;
Willy Tarreau79e57332018-10-02 16:01:16 +02002025 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02002026
Christopher Faulet6d1dd462019-07-15 14:36:03 +02002027 /* now retrieve the path */
2028 sl = http_get_stline(htx);
Amaury Denoyellec453f952021-07-06 11:40:12 +02002029 parser = http_uri_parser_init(htx_sl_req_uri(sl));
2030 path = http_parse_path(&parser);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02002031 if (path.len && *(path.ptr) == '/') {
2032 while (path.len--)
2033 hash = *(path.ptr++) + (hash << 6) + (hash << 16) - hash;
Willy Tarreau79e57332018-10-02 16:01:16 +02002034 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02002035
Willy Tarreau79e57332018-10-02 16:01:16 +02002036 hash = full_hash(hash);
2037
2038 smp->data.type = SMP_T_SINT;
2039 smp->data.u.sint = hash;
2040 smp->flags = SMP_F_VOL_1ST;
2041 return 1;
2042}
2043
2044/* This concatenates the source address with the 32-bit hash of the Host and
2045 * URL as returned by smp_fetch_base32(). The idea is to have per-source and
2046 * per-url counters. The result is a binary block from 8 to 20 bytes depending
2047 * on the source address length. The URL hash is stored before the address so
2048 * that in environments where IPv6 is insignificant, truncating the output to
2049 * 8 bytes would still work.
2050 */
2051static int smp_fetch_url32_src(const struct arg *args, struct sample *smp, const char *kw, void *private)
2052{
Willy Tarreaud68ff012022-05-27 08:57:21 +02002053 const struct sockaddr_storage *src = (smp->strm ? sc_src(smp->strm->scf) : NULL);
Willy Tarreau79e57332018-10-02 16:01:16 +02002054 struct buffer *temp;
Willy Tarreau79e57332018-10-02 16:01:16 +02002055
Christopher Faulet6fc817a2021-10-25 07:48:27 +02002056 if (!src)
Willy Tarreau79e57332018-10-02 16:01:16 +02002057 return 0;
2058
2059 if (!smp_fetch_url32(args, smp, kw, private))
2060 return 0;
2061
2062 temp = get_trash_chunk();
2063 *(unsigned int *) temp->area = htonl(smp->data.u.sint);
2064 temp->data += sizeof(unsigned int);
2065
Christopher Faulet6fc817a2021-10-25 07:48:27 +02002066 switch (src->ss_family) {
Willy Tarreau79e57332018-10-02 16:01:16 +02002067 case AF_INET:
2068 memcpy(temp->area + temp->data,
Christopher Faulet6fc817a2021-10-25 07:48:27 +02002069 &((struct sockaddr_in *)src)->sin_addr,
Willy Tarreau79e57332018-10-02 16:01:16 +02002070 4);
2071 temp->data += 4;
2072 break;
2073 case AF_INET6:
2074 memcpy(temp->area + temp->data,
Christopher Faulet6fc817a2021-10-25 07:48:27 +02002075 &((struct sockaddr_in6 *)src)->sin6_addr,
Willy Tarreau79e57332018-10-02 16:01:16 +02002076 16);
2077 temp->data += 16;
2078 break;
2079 default:
2080 return 0;
2081 }
2082
2083 smp->data.u.str = *temp;
2084 smp->data.type = SMP_T_BIN;
2085 return 1;
2086}
2087
2088/************************************************************************/
2089/* Other utility functions */
2090/************************************************************************/
2091
2092/* This function is used to validate the arguments passed to any "hdr" fetch
2093 * keyword. These keywords support an optional positive or negative occurrence
2094 * number. We must ensure that the number is greater than -MAX_HDR_HISTORY. It
2095 * is assumed that the types are already the correct ones. Returns 0 on error,
2096 * non-zero if OK. If <err> is not NULL, it will be filled with a pointer to an
2097 * error message in case of error, that the caller is responsible for freeing.
2098 * The initial location must either be freeable or NULL.
2099 * Note: this function's pointer is checked from Lua.
2100 */
2101int val_hdr(struct arg *arg, char **err_msg)
2102{
2103 if (arg && arg[1].type == ARGT_SINT && arg[1].data.sint < -MAX_HDR_HISTORY) {
2104 memprintf(err_msg, "header occurrence must be >= %d", -MAX_HDR_HISTORY);
2105 return 0;
2106 }
2107 return 1;
2108}
2109
2110/************************************************************************/
2111/* All supported sample fetch keywords must be declared here. */
2112/************************************************************************/
2113
2114/* Note: must not be declared <const> as its list will be overwritten */
2115static struct sample_fetch_kw_list sample_fetch_keywords = {ILH, {
2116 { "base", smp_fetch_base, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2117 { "base32", smp_fetch_base32, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2118 { "base32+src", smp_fetch_base32_src, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
Yves Lafonb4d37082021-02-11 11:01:28 +01002119 { "baseq", smp_fetch_base, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
Willy Tarreau79e57332018-10-02 16:01:16 +02002120
2121 /* capture are allocated and are permanent in the stream */
2122 { "capture.req.hdr", smp_fetch_capture_req_hdr, ARG1(1,SINT), NULL, SMP_T_STR, SMP_USE_HRQHP },
2123
2124 /* retrieve these captures from the HTTP logs */
2125 { "capture.req.method", smp_fetch_capture_req_method, 0, NULL, SMP_T_STR, SMP_USE_HRQHP },
2126 { "capture.req.uri", smp_fetch_capture_req_uri, 0, NULL, SMP_T_STR, SMP_USE_HRQHP },
2127 { "capture.req.ver", smp_fetch_capture_req_ver, 0, NULL, SMP_T_STR, SMP_USE_HRQHP },
2128
2129 { "capture.res.hdr", smp_fetch_capture_res_hdr, ARG1(1,SINT), NULL, SMP_T_STR, SMP_USE_HRSHP },
2130 { "capture.res.ver", smp_fetch_capture_res_ver, 0, NULL, SMP_T_STR, SMP_USE_HRQHP },
2131
2132 /* cookie is valid in both directions (eg: for "stick ...") but cook*
2133 * are only here to match the ACL's name, are request-only and are used
2134 * for ACL compatibility only.
2135 */
2136 { "cook", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
Christopher Fauletc1f40dd2019-05-16 10:07:30 +02002137 { "cookie", smp_fetch_chn_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRQHV|SMP_USE_HRSHV },
Willy Tarreau79e57332018-10-02 16:01:16 +02002138 { "cook_cnt", smp_fetch_cookie_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2139 { "cook_val", smp_fetch_cookie_val, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2140
2141 /* hdr is valid in both directions (eg: for "stick ...") but hdr_* are
2142 * only here to match the ACL's name, are request-only and are used for
2143 * ACL compatibility only.
2144 */
Christopher Fauletc1f40dd2019-05-16 10:07:30 +02002145 { "hdr", smp_fetch_chn_hdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRQHV|SMP_USE_HRSHV },
Willy Tarreau79e57332018-10-02 16:01:16 +02002146 { "hdr_cnt", smp_fetch_hdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2147 { "hdr_ip", smp_fetch_hdr_ip, ARG2(0,STR,SINT), val_hdr, SMP_T_IPV4, SMP_USE_HRQHV },
2148 { "hdr_val", smp_fetch_hdr_val, ARG2(0,STR,SINT), val_hdr, SMP_T_SINT, SMP_USE_HRQHV },
2149
Christopher Fauleta4063562019-08-02 11:51:37 +02002150 { "http_auth_type", smp_fetch_http_auth_type, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2151 { "http_auth_user", smp_fetch_http_auth_user, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2152 { "http_auth_pass", smp_fetch_http_auth_pass, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
Remi Tricot-Le Bretonf5dd3372021-10-01 15:36:53 +02002153 { "http_auth_bearer", smp_fetch_http_auth_bearer, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
Willy Tarreau79e57332018-10-02 16:01:16 +02002154 { "http_auth", smp_fetch_http_auth, ARG1(1,USR), NULL, SMP_T_BOOL, SMP_USE_HRQHV },
2155 { "http_auth_group", smp_fetch_http_auth_grp, ARG1(1,USR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2156 { "http_first_req", smp_fetch_http_first_req, 0, NULL, SMP_T_BOOL, SMP_USE_HRQHP },
2157 { "method", smp_fetch_meth, 0, NULL, SMP_T_METH, SMP_USE_HRQHP },
2158 { "path", smp_fetch_path, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
Christopher Faulete720c322020-09-02 17:25:18 +02002159 { "pathq", smp_fetch_path, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
Willy Tarreau79e57332018-10-02 16:01:16 +02002160 { "query", smp_fetch_query, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2161
2162 /* HTTP protocol on the request path */
2163 { "req.proto_http", smp_fetch_proto_http, 0, NULL, SMP_T_BOOL, SMP_USE_HRQHP },
2164 { "req_proto_http", smp_fetch_proto_http, 0, NULL, SMP_T_BOOL, SMP_USE_HRQHP },
2165
2166 /* HTTP version on the request path */
2167 { "req.ver", smp_fetch_rqver, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2168 { "req_ver", smp_fetch_rqver, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2169
2170 { "req.body", smp_fetch_body, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
2171 { "req.body_len", smp_fetch_body_len, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2172 { "req.body_size", smp_fetch_body_size, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2173 { "req.body_param", smp_fetch_body_param, ARG1(0,STR), NULL, SMP_T_BIN, SMP_USE_HRQHV },
2174
2175 { "req.hdrs", smp_fetch_hdrs, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
2176 { "req.hdrs_bin", smp_fetch_hdrs_bin, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
2177
2178 /* HTTP version on the response path */
2179 { "res.ver", smp_fetch_stver, 0, NULL, SMP_T_STR, SMP_USE_HRSHV },
2180 { "resp_ver", smp_fetch_stver, 0, NULL, SMP_T_STR, SMP_USE_HRSHV },
2181
Christopher Faulete596d182020-05-05 17:46:34 +02002182 { "res.body", smp_fetch_body, 0, NULL, SMP_T_BIN, SMP_USE_HRSHV },
2183 { "res.body_len", smp_fetch_body_len, 0, NULL, SMP_T_SINT, SMP_USE_HRSHV },
2184 { "res.body_size", smp_fetch_body_size, 0, NULL, SMP_T_SINT, SMP_USE_HRSHV },
2185
2186 { "res.hdrs", smp_fetch_hdrs, 0, NULL, SMP_T_BIN, SMP_USE_HRSHV },
2187 { "res.hdrs_bin", smp_fetch_hdrs_bin, 0, NULL, SMP_T_BIN, SMP_USE_HRSHV },
2188
Willy Tarreau79e57332018-10-02 16:01:16 +02002189 /* explicit req.{cook,hdr} are used to force the fetch direction to be request-only */
2190 { "req.cook", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2191 { "req.cook_cnt", smp_fetch_cookie_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2192 { "req.cook_val", smp_fetch_cookie_val, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2193
2194 { "req.fhdr", smp_fetch_fhdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRQHV },
2195 { "req.fhdr_cnt", smp_fetch_fhdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2196 { "req.hdr", smp_fetch_hdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRQHV },
2197 { "req.hdr_cnt", smp_fetch_hdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2198 { "req.hdr_ip", smp_fetch_hdr_ip, ARG2(0,STR,SINT), val_hdr, SMP_T_IPV4, SMP_USE_HRQHV },
2199 { "req.hdr_names", smp_fetch_hdr_names, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2200 { "req.hdr_val", smp_fetch_hdr_val, ARG2(0,STR,SINT), val_hdr, SMP_T_SINT, SMP_USE_HRQHV },
2201
2202 /* explicit req.{cook,hdr} are used to force the fetch direction to be response-only */
2203 { "res.cook", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRSHV },
2204 { "res.cook_cnt", smp_fetch_cookie_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2205 { "res.cook_val", smp_fetch_cookie_val, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2206
2207 { "res.fhdr", smp_fetch_fhdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRSHV },
2208 { "res.fhdr_cnt", smp_fetch_fhdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2209 { "res.hdr", smp_fetch_hdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRSHV },
2210 { "res.hdr_cnt", smp_fetch_hdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2211 { "res.hdr_ip", smp_fetch_hdr_ip, ARG2(0,STR,SINT), val_hdr, SMP_T_IPV4, SMP_USE_HRSHV },
2212 { "res.hdr_names", smp_fetch_hdr_names, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRSHV },
2213 { "res.hdr_val", smp_fetch_hdr_val, ARG2(0,STR,SINT), val_hdr, SMP_T_SINT, SMP_USE_HRSHV },
2214
2215 /* scook is valid only on the response and is used for ACL compatibility */
2216 { "scook", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRSHV },
2217 { "scook_cnt", smp_fetch_cookie_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2218 { "scook_val", smp_fetch_cookie_val, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
Willy Tarreau79e57332018-10-02 16:01:16 +02002219
2220 /* shdr is valid only on the response and is used for ACL compatibility */
2221 { "shdr", smp_fetch_hdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRSHV },
2222 { "shdr_cnt", smp_fetch_hdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2223 { "shdr_ip", smp_fetch_hdr_ip, ARG2(0,STR,SINT), val_hdr, SMP_T_IPV4, SMP_USE_HRSHV },
2224 { "shdr_val", smp_fetch_hdr_val, ARG2(0,STR,SINT), val_hdr, SMP_T_SINT, SMP_USE_HRSHV },
2225
2226 { "status", smp_fetch_stcode, 0, NULL, SMP_T_SINT, SMP_USE_HRSHP },
2227 { "unique-id", smp_fetch_uniqueid, 0, NULL, SMP_T_STR, SMP_SRC_L4SRV },
2228 { "url", smp_fetch_url, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2229 { "url32", smp_fetch_url32, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2230 { "url32+src", smp_fetch_url32_src, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
2231 { "url_ip", smp_fetch_url_ip, 0, NULL, SMP_T_IPV4, SMP_USE_HRQHV },
2232 { "url_port", smp_fetch_url_port, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2233 { "url_param", smp_fetch_url_param, ARG2(0,STR,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2234 { "urlp" , smp_fetch_url_param, ARG2(0,STR,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2235 { "urlp_val", smp_fetch_url_param_val, ARG2(0,STR,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
Christopher Faulet16032ab2020-04-30 11:30:00 +02002236
Willy Tarreau79e57332018-10-02 16:01:16 +02002237 { /* END */ },
2238}};
2239
Willy Tarreau0108d902018-11-25 19:14:37 +01002240INITCALL1(STG_REGISTER, sample_register_fetches, &sample_fetch_keywords);
Willy Tarreau79e57332018-10-02 16:01:16 +02002241
2242/*
2243 * Local variables:
2244 * c-indent-level: 8
2245 * c-basic-offset: 8
2246 * End:
2247 */