blob: 9cfcee2ab6c0312131cd980a992f1f557d2c40b0 [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
19#include <common/base64.h>
20#include <common/chunk.h>
21#include <common/compat.h>
22#include <common/config.h>
23#include <common/debug.h>
Willy Tarreauafba57a2018-12-11 13:44:24 +010024#include <common/h1.h>
Willy Tarreau79e57332018-10-02 16:01:16 +020025#include <common/http.h>
Willy Tarreaub96b77e2018-12-11 10:22:41 +010026#include <common/htx.h>
Willy Tarreau0108d902018-11-25 19:14:37 +010027#include <common/initcall.h>
Willy Tarreau79e57332018-10-02 16:01:16 +020028#include <common/memory.h>
29#include <common/standard.h>
30#include <common/version.h>
31
32#include <types/global.h>
33
34#include <proto/arg.h>
35#include <proto/auth.h>
Christopher Fauleteb2754b2019-07-16 14:49:01 +020036#include <proto/channel.h>
Willy Tarreau9a1efe12019-07-17 17:13:50 +020037#include <proto/connection.h>
Willy Tarreau79e57332018-10-02 16:01:16 +020038#include <proto/http_fetch.h>
Christopher Faulet53a899b2019-10-08 16:38:42 +020039#include <proto/h1_htx.h>
Christopher Fauletef453ed2018-10-24 21:39:27 +020040#include <proto/http_htx.h>
Willy Tarreau79e57332018-10-02 16:01:16 +020041#include <proto/log.h>
42#include <proto/obj_type.h>
Christopher Fauletfc9cfe42019-07-16 14:54:53 +020043#include <proto/http_ana.h>
Willy Tarreau79e57332018-10-02 16:01:16 +020044#include <proto/sample.h>
45#include <proto/stream.h>
46
47
48/* this struct is used between calls to smp_fetch_hdr() or smp_fetch_cookie() */
Christopher Fauletef453ed2018-10-24 21:39:27 +020049static THREAD_LOCAL struct http_hdr_ctx static_http_hdr_ctx;
Richard Russo458eafb2019-07-31 11:45:56 -070050/* this is used to convert raw connection buffers to htx */
51static THREAD_LOCAL struct buffer static_raw_htx_chunk;
52static THREAD_LOCAL char *static_raw_htx_buf;
Christopher Fauletef453ed2018-10-24 21:39:27 +020053
Christopher Faulet89dc4992019-04-17 12:02:59 +020054#define SMP_REQ_CHN(smp) (smp->strm ? &smp->strm->req : NULL)
55#define SMP_RES_CHN(smp) (smp->strm ? &smp->strm->res : NULL)
Willy Tarreau79e57332018-10-02 16:01:16 +020056
Richard Russo458eafb2019-07-31 11:45:56 -070057/* This function returns the static htx chunk, where raw connections get
58 * converted to HTX as needed for samplxsing.
59 */
60struct buffer *get_raw_htx_chunk(void)
61{
62 chunk_reset(&static_raw_htx_chunk);
63 return &static_raw_htx_chunk;
64}
65
66static int alloc_raw_htx_chunk_per_thread()
67{
68 static_raw_htx_buf = malloc(global.tune.bufsize);
69 if (!static_raw_htx_buf)
70 return 0;
71 chunk_init(&static_raw_htx_chunk, static_raw_htx_buf, global.tune.bufsize);
72 return 1;
73}
74
75static void free_raw_htx_chunk_per_thread()
76{
77 free(static_raw_htx_buf);
78 static_raw_htx_buf = NULL;
79}
80
81REGISTER_PER_THREAD_ALLOC(alloc_raw_htx_chunk_per_thread);
82REGISTER_PER_THREAD_FREE(free_raw_htx_chunk_per_thread);
83
Willy Tarreau79e57332018-10-02 16:01:16 +020084/*
85 * Returns the data from Authorization header. Function may be called more
86 * than once so data is stored in txn->auth_data. When no header is found
87 * or auth method is unknown auth_method is set to HTTP_AUTH_WRONG to avoid
88 * searching again for something we are unable to find anyway. However, if
89 * the result if valid, the cache is not reused because we would risk to
90 * have the credentials overwritten by another stream in parallel.
91 */
92
Christopher Fauletcd761952019-07-15 13:58:29 +020093static int get_http_auth(struct sample *smp, struct htx *htx)
Willy Tarreau79e57332018-10-02 16:01:16 +020094{
Christopher Faulet311c7ea2018-10-24 21:41:55 +020095 struct stream *s = smp->strm;
Willy Tarreau79e57332018-10-02 16:01:16 +020096 struct http_txn *txn = s->txn;
Christopher Faulet6d1dd462019-07-15 14:36:03 +020097 struct http_hdr_ctx ctx = { .blk = NULL };
98 struct ist hdr;
Willy Tarreau79e57332018-10-02 16:01:16 +020099 struct buffer auth_method;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200100 char *p;
Willy Tarreau79e57332018-10-02 16:01:16 +0200101 int len;
102
103#ifdef DEBUG_AUTH
104 printf("Auth for stream %p: %d\n", s, txn->auth.method);
105#endif
Willy Tarreau79e57332018-10-02 16:01:16 +0200106 if (txn->auth.method == HTTP_AUTH_WRONG)
107 return 0;
108
109 txn->auth.method = HTTP_AUTH_WRONG;
110
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200111 if (txn->flags & TX_USE_PX_CONN)
112 hdr = ist("Proxy-Authorization");
113 else
114 hdr = ist("Authorization");
Willy Tarreau79e57332018-10-02 16:01:16 +0200115
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200116 ctx.blk = NULL;
117 if (!http_find_header(htx, hdr, &ctx, 0))
118 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200119
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200120 p = memchr(ctx.value.ptr, ' ', ctx.value.len);
121 len = p - ctx.value.ptr;
122 if (!p || len <= 0)
123 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200124
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200125 if (chunk_initlen(&auth_method, ctx.value.ptr, 0, len) != 1)
126 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200127
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200128 chunk_initlen(&txn->auth.method_data, p + 1, 0, ctx.value.len - len - 1);
Willy Tarreau79e57332018-10-02 16:01:16 +0200129
130 if (!strncasecmp("Basic", auth_method.area, auth_method.data)) {
131 struct buffer *http_auth = get_trash_chunk();
132
133 len = base64dec(txn->auth.method_data.area,
134 txn->auth.method_data.data,
135 http_auth->area, global.tune.bufsize - 1);
136
137 if (len < 0)
138 return 0;
139
140
141 http_auth->area[len] = '\0';
142
143 p = strchr(http_auth->area, ':');
144
145 if (!p)
146 return 0;
147
148 txn->auth.user = http_auth->area;
149 *p = '\0';
150 txn->auth.pass = p+1;
151
152 txn->auth.method = HTTP_AUTH_BASIC;
153 return 1;
154 }
155
156 return 0;
157}
158
159/* This function ensures that the prerequisites for an L7 fetch are ready,
160 * which means that a request or response is ready. If some data is missing,
161 * a parsing attempt is made. This is useful in TCP-based ACLs which are able
Christopher Faulet5ec8bcb2019-04-17 12:04:12 +0200162 * to extract data from L7. If <vol> is non-null during a prefetch, another
163 * test is made to ensure the required information is not gone.
Christopher Fauletef453ed2018-10-24 21:39:27 +0200164 *
165 * The function returns :
166 * NULL with SMP_F_MAY_CHANGE in the sample flags if some data is missing to
167 * decide whether or not an HTTP message is present ;
168 * NULL if the requested data cannot be fetched or if it is certain that
169 * we'll never have any HTTP message there ;
170 * The HTX message if ready
171 */
Christopher Faulet5ec8bcb2019-04-17 12:04:12 +0200172struct htx *smp_prefetch_htx(struct sample *smp, struct channel *chn, int vol)
Christopher Fauletef453ed2018-10-24 21:39:27 +0200173{
Christopher Fauletef453ed2018-10-24 21:39:27 +0200174 struct stream *s = smp->strm;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200175 struct http_txn *txn = NULL;
176 struct htx *htx = NULL;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200177 struct http_msg *msg;
Christopher Fauletf1ba18d2018-11-26 21:37:08 +0100178 struct htx_sl *sl;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200179
180 /* Note: it is possible that <s> is NULL when called before stream
181 * initialization (eg: tcp-request connection), so this function is the
182 * one responsible for guarding against this case for all HTTP users.
183 */
Christopher Faulet89dc4992019-04-17 12:02:59 +0200184 if (!s || !chn)
Christopher Fauletef453ed2018-10-24 21:39:27 +0200185 return NULL;
186
187 if (!s->txn) {
188 if (unlikely(!http_alloc_txn(s)))
189 return NULL; /* not enough memory */
190 http_init_txn(s);
191 txn = s->txn;
192 }
Christopher Faulet89dc4992019-04-17 12:02:59 +0200193 txn = s->txn;
194 msg = (!(chn->flags & CF_ISRESP) ? &txn->req : &txn->rsp);
195 smp->data.type = SMP_T_BOOL;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200196
Christopher Fauleteca88542019-04-03 10:12:42 +0200197 if (IS_HTX_STRM(s)) {
Christopher Faulet89dc4992019-04-17 12:02:59 +0200198 htx = htxbuf(&chn->buf);
Christopher Fauletef453ed2018-10-24 21:39:27 +0200199
Christopher Faulet89dc4992019-04-17 12:02:59 +0200200 if (msg->msg_state == HTTP_MSG_ERROR || (htx->flags & HTX_FL_PARSING_ERROR))
201 return NULL;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200202
Christopher Faulet89dc4992019-04-17 12:02:59 +0200203 if (msg->msg_state < HTTP_MSG_BODY) {
204 /* Analyse not yet started */
Christopher Faulet29f17582019-05-23 11:03:26 +0200205 if (htx_is_empty(htx) || htx->first == -1) {
Christopher Fauletef453ed2018-10-24 21:39:27 +0200206 /* Parsing is done by the mux, just wait */
207 smp->flags |= SMP_F_MAY_CHANGE;
208 return NULL;
209 }
210 }
Christopher Faulet297fbb42019-05-13 14:41:27 +0200211 sl = http_get_stline(htx);
Christopher Faulet5ec8bcb2019-04-17 12:04:12 +0200212 if (vol && !sl) {
Christopher Faulet89dc4992019-04-17 12:02:59 +0200213 /* The start-line was already forwarded, it is too late to fetch anything */
214 return NULL;
215 }
Christopher Fauletef453ed2018-10-24 21:39:27 +0200216 }
Christopher Fauleteca88542019-04-03 10:12:42 +0200217 else { /* RAW mode */
Christopher Faulet89dc4992019-04-17 12:02:59 +0200218 struct buffer *buf;
219 struct h1m h1m;
Christopher Faulete4ab11b2019-06-11 15:05:37 +0200220 struct http_hdr hdrs[global.tune.max_http_hdr];
Christopher Faulet89dc4992019-04-17 12:02:59 +0200221 union h1_sl h1sl;
222 unsigned int flags = HTX_FL_NONE;
223 int ret;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200224
Christopher Faulet89dc4992019-04-17 12:02:59 +0200225 /* no HTTP fetch on the response in TCP mode */
226 if (chn->flags & CF_ISRESP)
227 return NULL;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200228
Christopher Faulet89dc4992019-04-17 12:02:59 +0200229 /* Now we are working on the request only */
230 buf = &chn->buf;
231 if (b_head(buf) + b_data(buf) > b_wrap(buf))
232 b_slow_realign(buf, trash.area, 0);
Christopher Fauletef453ed2018-10-24 21:39:27 +0200233
Christopher Faulet89dc4992019-04-17 12:02:59 +0200234 h1m_init_req(&h1m);
235 ret = h1_headers_to_hdr_list(b_head(buf), b_stop(buf),
236 hdrs, sizeof(hdrs)/sizeof(hdrs[0]), &h1m, &h1sl);
237 if (ret <= 0) {
238 /* Invalid or too big*/
239 if (ret < 0 || channel_full(&s->req, global.tune.maxrewrite))
Christopher Fauletef453ed2018-10-24 21:39:27 +0200240 return NULL;
Christopher Fauletf1ba18d2018-11-26 21:37:08 +0100241
Christopher Faulet89dc4992019-04-17 12:02:59 +0200242 /* wait for a full request */
243 smp->flags |= SMP_F_MAY_CHANGE;
244 return NULL;
245 }
Christopher Fauletf1ba18d2018-11-26 21:37:08 +0100246
Christopher Faulet89dc4992019-04-17 12:02:59 +0200247 /* OK we just got a valid HTTP mesage. We have to convert it
248 * into an HTX message.
249 */
250 if (unlikely(h1sl.rq.v.len == 0)) {
251 /* try to convert HTTP/0.9 requests to HTTP/1.0 */
252 if (h1sl.rq.meth != HTTP_METH_GET || !h1sl.rq.u.len)
Christopher Fauletef453ed2018-10-24 21:39:27 +0200253 return NULL;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200254 h1sl.rq.v = ist("HTTP/1.0");
Christopher Fauletef453ed2018-10-24 21:39:27 +0200255 }
Christopher Faulet89dc4992019-04-17 12:02:59 +0200256
257 /* Set HTX start-line flags */
258 if (h1m.flags & H1_MF_VER_11)
259 flags |= HTX_SL_F_VER_11;
260 if (h1m.flags & H1_MF_XFER_ENC)
261 flags |= HTX_SL_F_XFER_ENC;
262 flags |= HTX_SL_F_XFER_LEN;
263 if (h1m.flags & H1_MF_CHNK)
264 flags |= HTX_SL_F_CHNK;
265 else if (h1m.flags & H1_MF_CLEN)
266 flags |= HTX_SL_F_CLEN;
267
Richard Russo458eafb2019-07-31 11:45:56 -0700268 htx = htx_from_buf(get_raw_htx_chunk());
Christopher Faulet89dc4992019-04-17 12:02:59 +0200269 sl = htx_add_stline(htx, HTX_BLK_REQ_SL, flags, h1sl.rq.m, h1sl.rq.u, h1sl.rq.v);
270 if (!sl || !htx_add_all_headers(htx, hdrs))
Christopher Fauletef453ed2018-10-24 21:39:27 +0200271 return NULL;
Willy Tarreauce9bbf52019-05-13 08:32:31 +0200272 sl->info.req.meth = h1sl.rq.meth;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200273 }
274
275 /* OK we just got a valid HTTP message. If not already done by
276 * HTTP analyzers, we have some minor preparation to perform so
277 * that further checks can rely on HTTP tests.
278 */
279 if (sl && msg->msg_state < HTTP_MSG_BODY) {
280 if (!(chn->flags & CF_ISRESP)) {
281 txn->meth = sl->info.req.meth;
282 if (txn->meth == HTTP_METH_GET || txn->meth == HTTP_METH_HEAD)
283 s->flags |= SF_REDIRECTABLE;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200284 }
Christopher Faulet89dc4992019-04-17 12:02:59 +0200285 else
286 txn->status = sl->info.res.status;
287 if (sl->flags & HTX_SL_F_VER_11)
288 msg->flags |= HTTP_MSGF_VER_11;
Christopher Fauletef453ed2018-10-24 21:39:27 +0200289 }
290
291 /* everything's OK */
292 smp->data.u.sint = 1;
293 return htx;
294}
295
Willy Tarreau79e57332018-10-02 16:01:16 +0200296/* This function fetches the method of current HTTP request and stores
297 * it in the global pattern struct as a chunk. There are two possibilities :
298 * - if the method is known (not HTTP_METH_OTHER), its identifier is stored
299 * in <len> and <ptr> is NULL ;
300 * - if the method is unknown (HTTP_METH_OTHER), <ptr> points to the text and
301 * <len> to its length.
302 * This is intended to be used with pat_match_meth() only.
303 */
304static int smp_fetch_meth(const struct arg *args, struct sample *smp, const char *kw, void *private)
305{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200306 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200307 struct htx *htx = smp_prefetch_htx(smp, chn, 0);
Willy Tarreau79e57332018-10-02 16:01:16 +0200308 struct http_txn *txn;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200309 int meth;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200310
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200311 if (!htx)
312 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200313
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200314 txn = smp->strm->txn;
315 meth = txn->meth;
316 smp->data.type = SMP_T_METH;
317 smp->data.u.meth.meth = meth;
318 if (meth == HTTP_METH_OTHER) {
319 struct htx_sl *sl;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200320
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200321 if ((smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_RES) {
322 /* ensure the indexes are not affected */
323 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200324 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200325 sl = http_get_stline(htx);
326 smp->flags |= SMP_F_CONST;
327 smp->data.u.meth.str.area = HTX_SL_REQ_MPTR(sl);
328 smp->data.u.meth.str.data = HTX_SL_REQ_MLEN(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +0200329 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200330 smp->flags |= SMP_F_VOL_1ST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200331 return 1;
332}
333
334static int smp_fetch_rqver(const struct arg *args, struct sample *smp, const char *kw, void *private)
335{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200336 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200337 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
338 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200339 char *ptr;
340 int len;
341
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200342 if (!htx)
343 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200344
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200345 sl = http_get_stline(htx);
346 len = HTX_SL_REQ_VLEN(sl);
347 ptr = HTX_SL_REQ_VPTR(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +0200348
349 while ((len-- > 0) && (*ptr++ != '/'));
350 if (len <= 0)
351 return 0;
352
353 smp->data.type = SMP_T_STR;
354 smp->data.u.str.area = ptr;
355 smp->data.u.str.data = len;
356
357 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
358 return 1;
359}
360
361static int smp_fetch_stver(const struct arg *args, struct sample *smp, const char *kw, void *private)
362{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200363 struct channel *chn = SMP_RES_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200364 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
365 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200366 char *ptr;
367 int len;
368
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200369 if (!htx)
370 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200371
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200372 sl = http_get_stline(htx);
373 len = HTX_SL_RES_VLEN(sl);
374 ptr = HTX_SL_RES_VPTR(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +0200375
376 while ((len-- > 0) && (*ptr++ != '/'));
377 if (len <= 0)
378 return 0;
379
380 smp->data.type = SMP_T_STR;
381 smp->data.u.str.area = ptr;
382 smp->data.u.str.data = len;
383
384 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
385 return 1;
386}
387
388/* 3. Check on Status Code. We manipulate integers here. */
389static int smp_fetch_stcode(const struct arg *args, struct sample *smp, const char *kw, void *private)
390{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200391 struct channel *chn = SMP_RES_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200392 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
393 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200394 char *ptr;
395 int len;
396
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200397 if (!htx)
398 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200399
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200400 sl = http_get_stline(htx);
401 len = HTX_SL_RES_CLEN(sl);
402 ptr = HTX_SL_RES_CPTR(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +0200403
404 smp->data.type = SMP_T_SINT;
405 smp->data.u.sint = __strl2ui(ptr, len);
406 smp->flags = SMP_F_VOL_1ST;
407 return 1;
408}
409
410static int smp_fetch_uniqueid(const struct arg *args, struct sample *smp, const char *kw, void *private)
411{
Tim Duesterhusa17e6622020-03-05 20:19:02 +0100412 struct ist unique_id;
Tim Duesterhus2825b4b2020-02-28 15:13:34 +0100413
Willy Tarreau79e57332018-10-02 16:01:16 +0200414 if (LIST_ISEMPTY(&smp->sess->fe->format_unique_id))
415 return 0;
416
Tim Duesterhusa17e6622020-03-05 20:19:02 +0100417 unique_id = stream_generate_unique_id(smp->strm, &smp->sess->fe->format_unique_id);
418 if (!isttest(unique_id))
Tim Duesterhus2825b4b2020-02-28 15:13:34 +0100419 return 0;
420
Tim Duesterhusa17e6622020-03-05 20:19:02 +0100421 smp->data.u.str.area = smp->strm->unique_id.ptr;
422 smp->data.u.str.data = smp->strm->unique_id.len;
Tim Duesterhus2825b4b2020-02-28 15:13:34 +0100423 smp->data.type = SMP_T_STR;
Willy Tarreau79e57332018-10-02 16:01:16 +0200424 smp->flags = SMP_F_CONST;
425 return 1;
426}
427
428/* Returns a string block containing all headers including the
Joseph Herlant942eea32018-11-15 13:57:22 -0800429 * empty line which separes headers from the body. This is useful
430 * for some headers analysis.
Willy Tarreau79e57332018-10-02 16:01:16 +0200431 */
432static int smp_fetch_hdrs(const struct arg *args, struct sample *smp, const char *kw, void *private)
433{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200434 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200435 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
436 struct buffer *temp;
437 int32_t pos;
Willy Tarreau79e57332018-10-02 16:01:16 +0200438
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200439 if (!htx)
440 return 0;
441 temp = get_trash_chunk();
442 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
443 struct htx_blk *blk = htx_get_blk(htx, pos);
444 enum htx_blk_type type = htx_get_blk_type(blk);
Willy Tarreau79e57332018-10-02 16:01:16 +0200445
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200446 if (type == HTX_BLK_HDR) {
447 struct ist n = htx_get_blk_name(htx, blk);
448 struct ist v = htx_get_blk_value(htx, blk);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200449
Christopher Faulet53a899b2019-10-08 16:38:42 +0200450 if (!h1_format_htx_hdr(n, v, temp))
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200451 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200452 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200453 else if (type == HTX_BLK_EOH) {
454 if (!chunk_memcat(temp, "\r\n", 2))
455 return 0;
456 break;
457 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200458 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200459 smp->data.type = SMP_T_STR;
460 smp->data.u.str = *temp;
Willy Tarreau79e57332018-10-02 16:01:16 +0200461 return 1;
462}
463
464/* Returns the header request in a length/value encoded format.
465 * This is useful for exchanges with the SPOE.
466 *
467 * A "length value" is a multibyte code encoding numbers. It uses the
468 * SPOE format. The encoding is the following:
469 *
470 * Each couple "header name" / "header value" is composed
471 * like this:
472 * "length value" "header name bytes"
473 * "length value" "header value bytes"
474 * When the last header is reached, the header name and the header
475 * value are empty. Their length are 0
476 */
477static int smp_fetch_hdrs_bin(const struct arg *args, struct sample *smp, const char *kw, void *private)
478{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200479 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200480 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200481 struct buffer *temp;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200482 char *p, *end;
483 int32_t pos;
484 int ret;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200485
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200486 if (!htx)
487 return 0;
488 temp = get_trash_chunk();
489 p = temp->area;
490 end = temp->area + temp->size;
491 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
492 struct htx_blk *blk = htx_get_blk(htx, pos);
493 enum htx_blk_type type = htx_get_blk_type(blk);
494 struct ist n, v;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200495
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200496 if (type == HTX_BLK_HDR) {
497 n = htx_get_blk_name(htx,blk);
498 v = htx_get_blk_value(htx, blk);
Willy Tarreau79e57332018-10-02 16:01:16 +0200499
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200500 /* encode the header name. */
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200501 ret = encode_varint(n.len, &p, end);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200502 if (ret == -1)
503 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200504 if (p + n.len > end)
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200505 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200506 memcpy(p, n.ptr, n.len);
507 p += n.len;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200508
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200509 /* encode the header value. */
510 ret = encode_varint(v.len, &p, end);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200511 if (ret == -1)
512 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200513 if (p + v.len > end)
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200514 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200515 memcpy(p, v.ptr, v.len);
516 p += v.len;
Willy Tarreau79e57332018-10-02 16:01:16 +0200517
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200518 }
519 else if (type == HTX_BLK_EOH) {
520 /* encode the end of the header list with empty
521 * header name and header value.
522 */
523 ret = encode_varint(0, &p, end);
524 if (ret == -1)
525 return 0;
526 ret = encode_varint(0, &p, end);
527 if (ret == -1)
528 return 0;
529 break;
530 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200531 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200532
533 /* Initialise sample data which will be filled. */
534 smp->data.type = SMP_T_BIN;
535 smp->data.u.str.area = temp->area;
536 smp->data.u.str.data = p - temp->area;
537 smp->data.u.str.size = temp->size;
Willy Tarreau79e57332018-10-02 16:01:16 +0200538 return 1;
539}
540
541/* returns the longest available part of the body. This requires that the body
542 * has been waited for using http-buffer-request.
543 */
544static int smp_fetch_body(const struct arg *args, struct sample *smp, const char *kw, void *private)
545{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200546 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200547 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
Willy Tarreau79e57332018-10-02 16:01:16 +0200548 struct buffer *temp;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200549 int32_t pos;
Willy Tarreau79e57332018-10-02 16:01:16 +0200550
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200551 if (!htx)
552 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200553
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200554 temp = get_trash_chunk();
555 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
556 struct htx_blk *blk = htx_get_blk(htx, pos);
557 enum htx_blk_type type = htx_get_blk_type(blk);
Willy Tarreau79e57332018-10-02 16:01:16 +0200558
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200559 if (type == HTX_BLK_EOM || type == HTX_BLK_TLR || type == HTX_BLK_EOT)
560 break;
561 if (type == HTX_BLK_DATA) {
Christopher Faulet53a899b2019-10-08 16:38:42 +0200562 if (!h1_format_htx_data(htx_get_blk_value(htx, blk), temp, 0))
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200563 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200564 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200565 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200566
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200567 smp->data.type = SMP_T_BIN;
568 smp->data.u.str = *temp;
569 smp->flags = SMP_F_VOL_TEST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200570 return 1;
571}
572
573
574/* returns the available length of the body. This requires that the body
575 * has been waited for using http-buffer-request.
576 */
577static int smp_fetch_body_len(const struct arg *args, struct sample *smp, const char *kw, void *private)
578{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200579 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200580 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
581 int32_t pos;
582 unsigned long long len = 0;
Christopher Fauletc16317d2018-12-12 14:11:22 +0100583
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200584 if (!htx)
585 return 0;
Christopher Fauletc16317d2018-12-12 14:11:22 +0100586
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200587 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
588 struct htx_blk *blk = htx_get_blk(htx, pos);
589 enum htx_blk_type type = htx_get_blk_type(blk);
Christopher Fauletc16317d2018-12-12 14:11:22 +0100590
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200591 if (type == HTX_BLK_EOM || type == HTX_BLK_TLR || type == HTX_BLK_EOT)
592 break;
593 if (type == HTX_BLK_DATA)
594 len += htx_get_blksz(blk);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200595 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200596
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200597 smp->data.type = SMP_T_SINT;
598 smp->data.u.sint = len;
599 smp->flags = SMP_F_VOL_TEST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200600 return 1;
601}
602
603
604/* returns the advertised length of the body, or the advertised size of the
605 * chunks available in the buffer. This requires that the body has been waited
606 * for using http-buffer-request.
607 */
608static int smp_fetch_body_size(const struct arg *args, struct sample *smp, const char *kw, void *private)
609{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200610 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200611 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
612 int32_t pos;
613 unsigned long long len = 0;
Christopher Faulet89dc4992019-04-17 12:02:59 +0200614
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200615 if (!htx)
616 return 0;
Christopher Fauletc16317d2018-12-12 14:11:22 +0100617
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200618 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
619 struct htx_blk *blk = htx_get_blk(htx, pos);
620 enum htx_blk_type type = htx_get_blk_type(blk);
Christopher Fauletc16317d2018-12-12 14:11:22 +0100621
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200622 if (type == HTX_BLK_EOM || type == HTX_BLK_TLR || type == HTX_BLK_EOT)
623 break;
624 if (type == HTX_BLK_DATA)
625 len += htx_get_blksz(blk);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200626 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200627 if (htx->extra != ULLONG_MAX)
628 len += htx->extra;
Willy Tarreau79e57332018-10-02 16:01:16 +0200629
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200630 smp->data.type = SMP_T_SINT;
631 smp->data.u.sint = len;
632 smp->flags = SMP_F_VOL_TEST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200633 return 1;
634}
635
636
637/* 4. Check on URL/URI. A pointer to the URI is stored. */
638static int smp_fetch_url(const struct arg *args, struct sample *smp, const char *kw, void *private)
639{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200640 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200641 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
642 struct htx_sl *sl;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200643
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200644 if (!htx)
645 return 0;
646 sl = http_get_stline(htx);
647 smp->data.type = SMP_T_STR;
648 smp->data.u.str.area = HTX_SL_REQ_UPTR(sl);
649 smp->data.u.str.data = HTX_SL_REQ_ULEN(sl);
650 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200651 return 1;
652}
653
654static int smp_fetch_url_ip(const struct arg *args, struct sample *smp, const char *kw, void *private)
655{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200656 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200657 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
658 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200659 struct sockaddr_storage addr;
660
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200661 if (!htx)
662 return 0;
663 sl = http_get_stline(htx);
664 url2sa(HTX_SL_REQ_UPTR(sl), HTX_SL_REQ_ULEN(sl), &addr, NULL);
Willy Tarreau79e57332018-10-02 16:01:16 +0200665
Willy Tarreau79e57332018-10-02 16:01:16 +0200666 if (((struct sockaddr_in *)&addr)->sin_family != AF_INET)
667 return 0;
668
669 smp->data.type = SMP_T_IPV4;
670 smp->data.u.ipv4 = ((struct sockaddr_in *)&addr)->sin_addr;
671 smp->flags = 0;
672 return 1;
673}
674
675static int smp_fetch_url_port(const struct arg *args, struct sample *smp, const char *kw, void *private)
676{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200677 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200678 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
679 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +0200680 struct sockaddr_storage addr;
681
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200682 if (!htx)
683 return 0;
684 sl = http_get_stline(htx);
685 url2sa(HTX_SL_REQ_UPTR(sl), HTX_SL_REQ_ULEN(sl), &addr, NULL);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200686
Willy Tarreau79e57332018-10-02 16:01:16 +0200687 if (((struct sockaddr_in *)&addr)->sin_family != AF_INET)
688 return 0;
689
690 smp->data.type = SMP_T_SINT;
691 smp->data.u.sint = ntohs(((struct sockaddr_in *)&addr)->sin_port);
692 smp->flags = 0;
693 return 1;
694}
695
696/* Fetch an HTTP header. A pointer to the beginning of the value is returned.
697 * Accepts an optional argument of type string containing the header field name,
698 * and an optional argument of type signed or unsigned integer to request an
699 * explicit occurrence of the header. Note that in the event of a missing name,
700 * headers are considered from the first one. It does not stop on commas and
701 * returns full lines instead (useful for User-Agent or Date for example).
702 */
703static int smp_fetch_fhdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
704{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200705 /* possible keywords: req.fhdr, res.fhdr */
706 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200707 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
708 struct http_hdr_ctx *ctx = smp->ctx.a[0];
709 struct ist name;
Willy Tarreau79e57332018-10-02 16:01:16 +0200710 int occ = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200711
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200712 if (!ctx) {
713 /* first call */
714 ctx = &static_http_hdr_ctx;
715 ctx->blk = NULL;
716 smp->ctx.a[0] = ctx;
717 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200718
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200719 if (args) {
720 if (args[0].type != ARGT_STR)
Willy Tarreau79e57332018-10-02 16:01:16 +0200721 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200722 name.ptr = args[0].data.str.area;
723 name.len = args[0].data.str.data;
Willy Tarreau79e57332018-10-02 16:01:16 +0200724
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200725 if (args[1].type == ARGT_SINT)
726 occ = args[1].data.sint;
Willy Tarreau79e57332018-10-02 16:01:16 +0200727 }
728
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200729 if (!htx)
730 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200731
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200732 if (ctx && !(smp->flags & SMP_F_NOT_LAST))
733 /* search for header from the beginning */
734 ctx->blk = NULL;
Willy Tarreau79e57332018-10-02 16:01:16 +0200735
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200736 if (!occ && !(smp->opt & SMP_OPT_ITERATE))
737 /* no explicit occurrence and single fetch => last header by default */
738 occ = -1;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200739
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200740 if (!occ)
741 /* prepare to report multiple occurrences for ACL fetches */
742 smp->flags |= SMP_F_NOT_LAST;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200743
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200744 smp->data.type = SMP_T_STR;
745 smp->flags |= SMP_F_VOL_HDR | SMP_F_CONST;
746 if (http_get_htx_fhdr(htx, name, occ, ctx, &smp->data.u.str.area, &smp->data.u.str.data))
747 return 1;
Willy Tarreau79e57332018-10-02 16:01:16 +0200748 smp->flags &= ~SMP_F_NOT_LAST;
749 return 0;
750}
751
752/* 6. Check on HTTP header count. The number of occurrences is returned.
753 * Accepts exactly 1 argument of type string. It does not stop on commas and
754 * returns full lines instead (useful for User-Agent or Date for example).
755 */
756static int smp_fetch_fhdr_cnt(const struct arg *args, struct sample *smp, const char *kw, void *private)
757{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200758 /* possible keywords: req.fhdr_cnt, res.fhdr_cnt */
759 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200760 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
761 struct http_hdr_ctx ctx;
762 struct ist name;
Willy Tarreau79e57332018-10-02 16:01:16 +0200763 int cnt;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200764
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200765 if (!htx)
766 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200767
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200768 if (args && args->type == ARGT_STR) {
769 name.ptr = args->data.str.area;
770 name.len = args->data.str.data;
771 } else {
772 name.ptr = NULL;
773 name.len = 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200774 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200775
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200776 ctx.blk = NULL;
777 cnt = 0;
778 while (http_find_header(htx, name, &ctx, 1))
779 cnt++;
Willy Tarreau79e57332018-10-02 16:01:16 +0200780 smp->data.type = SMP_T_SINT;
781 smp->data.u.sint = cnt;
782 smp->flags = SMP_F_VOL_HDR;
783 return 1;
784}
785
786static int smp_fetch_hdr_names(const struct arg *args, struct sample *smp, const char *kw, void *private)
787{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200788 /* possible keywords: req.hdr_names, res.hdr_names */
789 struct channel *chn = ((kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200790 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
Willy Tarreau79e57332018-10-02 16:01:16 +0200791 struct buffer *temp;
792 char del = ',';
793
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200794 int32_t pos;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200795
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200796 if (!htx)
797 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200798
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200799 if (args && args->type == ARGT_STR)
800 del = *args[0].data.str.area;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200801
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200802 temp = get_trash_chunk();
803 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
804 struct htx_blk *blk = htx_get_blk(htx, pos);
805 enum htx_blk_type type = htx_get_blk_type(blk);
806 struct ist n;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200807
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200808 if (type == HTX_BLK_EOH)
809 break;
810 if (type != HTX_BLK_HDR)
811 continue;
812 n = htx_get_blk_name(htx, blk);
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200813
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200814 if (temp->data)
815 temp->area[temp->data++] = del;
816 chunk_memcat(temp, n.ptr, n.len);
Willy Tarreau79e57332018-10-02 16:01:16 +0200817 }
818
819 smp->data.type = SMP_T_STR;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200820 smp->data.u.str = *temp;
Willy Tarreau79e57332018-10-02 16:01:16 +0200821 smp->flags = SMP_F_VOL_HDR;
822 return 1;
823}
824
825/* Fetch an HTTP header. A pointer to the beginning of the value is returned.
826 * Accepts an optional argument of type string containing the header field name,
827 * and an optional argument of type signed or unsigned integer to request an
828 * explicit occurrence of the header. Note that in the event of a missing name,
829 * headers are considered from the first one.
830 */
831static int smp_fetch_hdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
832{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200833 /* possible keywords: req.hdr / hdr, res.hdr / shdr */
834 struct channel *chn = ((kw[0] == 'h' || kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200835 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
836 struct http_hdr_ctx *ctx = smp->ctx.a[0];
837 struct ist name;
Willy Tarreau79e57332018-10-02 16:01:16 +0200838 int occ = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200839
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200840 if (!ctx) {
841 /* first call */
842 ctx = &static_http_hdr_ctx;
843 ctx->blk = NULL;
844 smp->ctx.a[0] = ctx;
845 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200846
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200847 if (args) {
848 if (args[0].type != ARGT_STR)
Willy Tarreau79e57332018-10-02 16:01:16 +0200849 return 0;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200850 name.ptr = args[0].data.str.area;
851 name.len = args[0].data.str.data;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200852
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200853 if (args[1].type == ARGT_SINT)
854 occ = args[1].data.sint;
Willy Tarreau79e57332018-10-02 16:01:16 +0200855 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200856
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200857 if (!htx)
858 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200859
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200860 if (ctx && !(smp->flags & SMP_F_NOT_LAST))
861 /* search for header from the beginning */
862 ctx->blk = NULL;
Willy Tarreau79e57332018-10-02 16:01:16 +0200863
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200864 if (!occ && !(smp->opt & SMP_OPT_ITERATE))
865 /* no explicit occurrence and single fetch => last header by default */
866 occ = -1;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200867
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200868 if (!occ)
869 /* prepare to report multiple occurrences for ACL fetches */
870 smp->flags |= SMP_F_NOT_LAST;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200871
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200872 smp->data.type = SMP_T_STR;
873 smp->flags |= SMP_F_VOL_HDR | SMP_F_CONST;
874 if (http_get_htx_hdr(htx, name, occ, ctx, &smp->data.u.str.area, &smp->data.u.str.data))
875 return 1;
Willy Tarreau79e57332018-10-02 16:01:16 +0200876
877 smp->flags &= ~SMP_F_NOT_LAST;
878 return 0;
879}
880
Christopher Fauletc1f40dd2019-05-16 10:07:30 +0200881/* Same than smp_fetch_hdr() but only relies on the sample direction to choose
882 * the right channel. So instead of duplicating the code, we just change the
883 * keyword and then fallback on smp_fetch_hdr().
884 */
885static int smp_fetch_chn_hdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
886{
887 kw = ((smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ ? "req.hdr" : "res.hdr");
888 return smp_fetch_hdr(args, smp, kw, private);
889}
890
Willy Tarreau79e57332018-10-02 16:01:16 +0200891/* 6. Check on HTTP header count. The number of occurrences is returned.
892 * Accepts exactly 1 argument of type string.
893 */
894static int smp_fetch_hdr_cnt(const struct arg *args, struct sample *smp, const char *kw, void *private)
895{
Christopher Faulet89dc4992019-04-17 12:02:59 +0200896 /* possible keywords: req.hdr_cnt / hdr_cnt, res.hdr_cnt / shdr_cnt */
897 struct channel *chn = ((kw[0] == 'h' || kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200898 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
899 struct http_hdr_ctx ctx;
900 struct ist name;
Willy Tarreau79e57332018-10-02 16:01:16 +0200901 int cnt;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200902
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200903 if (!htx)
904 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200905
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200906 if (args && args->type == ARGT_STR) {
907 name.ptr = args->data.str.area;
908 name.len = args->data.str.data;
909 } else {
910 name.ptr = NULL;
911 name.len = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +0200912 }
Willy Tarreau79e57332018-10-02 16:01:16 +0200913
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200914 ctx.blk = NULL;
915 cnt = 0;
916 while (http_find_header(htx, name, &ctx, 0))
917 cnt++;
Willy Tarreau79e57332018-10-02 16:01:16 +0200918
919 smp->data.type = SMP_T_SINT;
920 smp->data.u.sint = cnt;
921 smp->flags = SMP_F_VOL_HDR;
922 return 1;
923}
924
925/* Fetch an HTTP header's integer value. The integer value is returned. It
926 * takes a mandatory argument of type string and an optional one of type int
927 * to designate a specific occurrence. It returns an unsigned integer, which
928 * may or may not be appropriate for everything.
929 */
930static int smp_fetch_hdr_val(const struct arg *args, struct sample *smp, const char *kw, void *private)
931{
932 int ret = smp_fetch_hdr(args, smp, kw, private);
933
934 if (ret > 0) {
935 smp->data.type = SMP_T_SINT;
936 smp->data.u.sint = strl2ic(smp->data.u.str.area,
937 smp->data.u.str.data);
938 }
939
940 return ret;
941}
942
943/* Fetch an HTTP header's IP value. takes a mandatory argument of type string
944 * and an optional one of type int to designate a specific occurrence.
945 * It returns an IPv4 or IPv6 address.
946 */
947static int smp_fetch_hdr_ip(const struct arg *args, struct sample *smp, const char *kw, void *private)
948{
949 int ret;
950
951 while ((ret = smp_fetch_hdr(args, smp, kw, private)) > 0) {
952 if (url2ipv4((char *) smp->data.u.str.area, &smp->data.u.ipv4)) {
953 smp->data.type = SMP_T_IPV4;
954 break;
955 } else {
956 struct buffer *temp = get_trash_chunk();
957 if (smp->data.u.str.data < temp->size - 1) {
958 memcpy(temp->area, smp->data.u.str.area,
959 smp->data.u.str.data);
960 temp->area[smp->data.u.str.data] = '\0';
961 if (inet_pton(AF_INET6, temp->area, &smp->data.u.ipv6)) {
962 smp->data.type = SMP_T_IPV6;
963 break;
964 }
965 }
966 }
967
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200968 /* if the header doesn't match an IP address, fetch next one */
969 if (!(smp->flags & SMP_F_NOT_LAST))
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200970 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200971 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200972 return ret;
973}
Willy Tarreau79e57332018-10-02 16:01:16 +0200974
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200975/* 8. Check on URI PATH. A pointer to the PATH is stored. The path starts at
976 * the first '/' after the possible hostname, and ends before the possible '?'.
977 */
978static int smp_fetch_path(const struct arg *args, struct sample *smp, const char *kw, void *private)
979{
980 struct channel *chn = SMP_REQ_CHN(smp);
981 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
982 struct htx_sl *sl;
983 struct ist path;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200984
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200985 if (!htx)
986 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200987
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200988 sl = http_get_stline(htx);
Jerome Magnin4fb196c2020-02-21 10:49:12 +0100989 path = iststop(http_get_path(htx_sl_req_uri(sl)), '?');
Tim Duesterhused526372020-03-05 17:56:33 +0100990 if (!isttest(path))
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200991 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +0200992
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200993 /* OK, we got the '/' ! */
994 smp->data.type = SMP_T_STR;
995 smp->data.u.str.area = path.ptr;
Jerome Magnin4fb196c2020-02-21 10:49:12 +0100996 smp->data.u.str.data = path.len;
Christopher Faulet6d1dd462019-07-15 14:36:03 +0200997 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
Willy Tarreau79e57332018-10-02 16:01:16 +0200998 return 1;
999}
1000
1001/* This produces a concatenation of the first occurrence of the Host header
1002 * followed by the path component if it begins with a slash ('/'). This means
1003 * that '*' will not be added, resulting in exactly the first Host entry.
1004 * If no Host header is found, then the path is returned as-is. The returned
1005 * value is stored in the trash so it does not need to be marked constant.
1006 * The returned sample is of type string.
1007 */
1008static int smp_fetch_base(const struct arg *args, struct sample *smp, const char *kw, void *private)
1009{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001010 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001011 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1012 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +02001013 struct buffer *temp;
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001014 struct http_hdr_ctx ctx;
1015 struct ist path;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001016
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001017 if (!htx)
1018 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001019
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001020 ctx.blk = NULL;
1021 if (!http_find_header(htx, ist("Host"), &ctx, 0) || !ctx.value.len)
1022 return smp_fetch_path(args, smp, kw, private);
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001023
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001024 /* OK we have the header value in ctx.value */
1025 temp = get_trash_chunk();
1026 chunk_memcat(temp, ctx.value.ptr, ctx.value.len);
Willy Tarreau79e57332018-10-02 16:01:16 +02001027
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001028 /* now retrieve the path */
1029 sl = http_get_stline(htx);
1030 path = http_get_path(htx_sl_req_uri(sl));
Tim Duesterhused526372020-03-05 17:56:33 +01001031 if (isttest(path)) {
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001032 size_t len;
Willy Tarreau79e57332018-10-02 16:01:16 +02001033
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001034 for (len = 0; len < path.len && *(path.ptr + len) != '?'; len++)
1035 ;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001036
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001037 if (len && *(path.ptr) == '/')
1038 chunk_memcat(temp, path.ptr, len);
Willy Tarreau79e57332018-10-02 16:01:16 +02001039 }
1040
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001041 smp->data.type = SMP_T_STR;
1042 smp->data.u.str = *temp;
Willy Tarreau79e57332018-10-02 16:01:16 +02001043 smp->flags = SMP_F_VOL_1ST;
1044 return 1;
1045}
1046
1047/* This produces a 32-bit hash of the concatenation of the first occurrence of
1048 * the Host header followed by the path component if it begins with a slash ('/').
1049 * This means that '*' will not be added, resulting in exactly the first Host
1050 * entry. If no Host header is found, then the path is used. The resulting value
1051 * is hashed using the path hash followed by a full avalanche hash and provides a
1052 * 32-bit integer value. This fetch is useful for tracking per-path activity on
1053 * high-traffic sites without having to store whole paths.
1054 */
1055static int smp_fetch_base32(const struct arg *args, struct sample *smp, const char *kw, void *private)
1056{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001057 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001058 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1059 struct htx_sl *sl;
1060 struct http_hdr_ctx ctx;
1061 struct ist path;
Willy Tarreau79e57332018-10-02 16:01:16 +02001062 unsigned int hash = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001063
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001064 if (!htx)
1065 return 0;
Dragan Dosen8861e1c2019-02-12 19:50:31 +01001066
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001067 ctx.blk = NULL;
1068 if (http_find_header(htx, ist("Host"), &ctx, 0)) {
1069 /* OK we have the header value in ctx.value */
1070 while (ctx.value.len--)
1071 hash = *(ctx.value.ptr++) + (hash << 6) + (hash << 16) - hash;
Willy Tarreau79e57332018-10-02 16:01:16 +02001072 }
1073
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001074 /* now retrieve the path */
1075 sl = http_get_stline(htx);
1076 path = http_get_path(htx_sl_req_uri(sl));
Tim Duesterhused526372020-03-05 17:56:33 +01001077 if (isttest(path)) {
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001078 size_t len;
Willy Tarreau79e57332018-10-02 16:01:16 +02001079
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001080 for (len = 0; len < path.len && *(path.ptr + len) != '?'; len++)
1081 ;
Willy Tarreau79e57332018-10-02 16:01:16 +02001082
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001083 if (len && *(path.ptr) == '/') {
1084 while (len--)
1085 hash = *(path.ptr++) + (hash << 6) + (hash << 16) - hash;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001086 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001087 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001088
Willy Tarreau79e57332018-10-02 16:01:16 +02001089 hash = full_hash(hash);
1090
1091 smp->data.type = SMP_T_SINT;
1092 smp->data.u.sint = hash;
1093 smp->flags = SMP_F_VOL_1ST;
1094 return 1;
1095}
1096
1097/* This concatenates the source address with the 32-bit hash of the Host and
1098 * path as returned by smp_fetch_base32(). The idea is to have per-source and
1099 * per-path counters. The result is a binary block from 8 to 20 bytes depending
1100 * on the source address length. The path hash is stored before the address so
1101 * that in environments where IPv6 is insignificant, truncating the output to
1102 * 8 bytes would still work.
1103 */
1104static int smp_fetch_base32_src(const struct arg *args, struct sample *smp, const char *kw, void *private)
1105{
1106 struct buffer *temp;
1107 struct connection *cli_conn = objt_conn(smp->sess->origin);
1108
Willy Tarreaucd7ca792019-07-17 16:57:03 +02001109 if (!cli_conn || !conn_get_src(cli_conn))
Willy Tarreau79e57332018-10-02 16:01:16 +02001110 return 0;
1111
1112 if (!smp_fetch_base32(args, smp, kw, private))
1113 return 0;
1114
1115 temp = get_trash_chunk();
1116 *(unsigned int *) temp->area = htonl(smp->data.u.sint);
1117 temp->data += sizeof(unsigned int);
1118
Willy Tarreau9a1efe12019-07-17 17:13:50 +02001119 switch (cli_conn->src->ss_family) {
Willy Tarreau79e57332018-10-02 16:01:16 +02001120 case AF_INET:
1121 memcpy(temp->area + temp->data,
Willy Tarreau9a1efe12019-07-17 17:13:50 +02001122 &((struct sockaddr_in *)cli_conn->src)->sin_addr,
Willy Tarreau79e57332018-10-02 16:01:16 +02001123 4);
1124 temp->data += 4;
1125 break;
1126 case AF_INET6:
1127 memcpy(temp->area + temp->data,
Willy Tarreau9a1efe12019-07-17 17:13:50 +02001128 &((struct sockaddr_in6 *)cli_conn->src)->sin6_addr,
Willy Tarreau79e57332018-10-02 16:01:16 +02001129 16);
1130 temp->data += 16;
1131 break;
1132 default:
1133 return 0;
1134 }
1135
1136 smp->data.u.str = *temp;
1137 smp->data.type = SMP_T_BIN;
1138 return 1;
1139}
1140
1141/* Extracts the query string, which comes after the question mark '?'. If no
1142 * question mark is found, nothing is returned. Otherwise it returns a sample
1143 * of type string carrying the whole query string.
1144 */
1145static int smp_fetch_query(const struct arg *args, struct sample *smp, const char *kw, void *private)
1146{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001147 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001148 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1149 struct htx_sl *sl;
Willy Tarreau79e57332018-10-02 16:01:16 +02001150 char *ptr, *end;
1151
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001152 if (!htx)
1153 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001154
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001155 sl = http_get_stline(htx);
1156 ptr = HTX_SL_REQ_UPTR(sl);
1157 end = HTX_SL_REQ_UPTR(sl) + HTX_SL_REQ_ULEN(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +02001158
1159 /* look up the '?' */
1160 do {
1161 if (ptr == end)
1162 return 0;
1163 } while (*ptr++ != '?');
1164
1165 smp->data.type = SMP_T_STR;
1166 smp->data.u.str.area = ptr;
1167 smp->data.u.str.data = end - ptr;
1168 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
1169 return 1;
1170}
1171
1172static int smp_fetch_proto_http(const struct arg *args, struct sample *smp, const char *kw, void *private)
1173{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001174 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001175 struct htx *htx = smp_prefetch_htx(smp, chn, 0);
Willy Tarreau79e57332018-10-02 16:01:16 +02001176
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001177 if (!htx)
1178 return 0;
1179 smp->data.type = SMP_T_BOOL;
Willy Tarreau79e57332018-10-02 16:01:16 +02001180 smp->data.u.sint = 1;
1181 return 1;
1182}
1183
1184/* return a valid test if the current request is the first one on the connection */
1185static int smp_fetch_http_first_req(const struct arg *args, struct sample *smp, const char *kw, void *private)
1186{
1187 smp->data.type = SMP_T_BOOL;
1188 smp->data.u.sint = !(smp->strm->txn->flags & TX_NOT_FIRST);
1189 return 1;
1190}
1191
Christopher Fauleta4063562019-08-02 11:51:37 +02001192/* Fetch the authentication method if there is an Authorization header. It
1193 * relies on get_http_auth()
1194 */
1195static int smp_fetch_http_auth_type(const struct arg *args, struct sample *smp, const char *kw, void *private)
1196{
1197 struct channel *chn = SMP_REQ_CHN(smp);
1198 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1199 struct http_txn *txn;
1200
1201 if (!htx)
1202 return 0;
1203
1204 txn = smp->strm->txn;
1205 if (!get_http_auth(smp, htx))
1206 return 0;
1207
1208 switch (txn->auth.method) {
1209 case HTTP_AUTH_BASIC:
1210 smp->data.u.str.area = "Basic";
1211 smp->data.u.str.data = 5;
1212 break;
1213 case HTTP_AUTH_DIGEST:
1214 /* Unexpected because not supported */
1215 smp->data.u.str.area = "Digest";
1216 smp->data.u.str.data = 6;
1217 break;
1218 default:
1219 return 0;
1220 }
1221
1222 smp->data.type = SMP_T_STR;
1223 smp->flags = SMP_F_CONST;
1224 return 1;
1225}
1226
1227/* Fetch the user supplied if there is an Authorization header. It relies on
1228 * get_http_auth()
1229 */
1230static int smp_fetch_http_auth_user(const struct arg *args, struct sample *smp, const char *kw, void *private)
1231{
1232 struct channel *chn = SMP_REQ_CHN(smp);
1233 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1234 struct http_txn *txn;
1235
1236 if (!htx)
1237 return 0;
1238
1239 txn = smp->strm->txn;
1240 if (!get_http_auth(smp, htx))
1241 return 0;
1242
1243 smp->data.type = SMP_T_STR;
1244 smp->data.u.str.area = txn->auth.user;
1245 smp->data.u.str.data = strlen(txn->auth.user);
1246 smp->flags = SMP_F_CONST;
1247 return 1;
1248}
1249
1250/* Fetch the password supplied if there is an Authorization header. It relies on
1251 * get_http_auth()
1252 */
1253static int smp_fetch_http_auth_pass(const struct arg *args, struct sample *smp, const char *kw, void *private)
1254{
1255 struct channel *chn = SMP_REQ_CHN(smp);
1256 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1257 struct http_txn *txn;
1258
1259 if (!htx)
1260 return 0;
1261
1262 txn = smp->strm->txn;
1263 if (!get_http_auth(smp, htx))
1264 return 0;
1265
1266 smp->data.type = SMP_T_STR;
1267 smp->data.u.str.area = txn->auth.pass;
1268 smp->data.u.str.data = strlen(txn->auth.pass);
1269 smp->flags = SMP_F_CONST;
1270 return 1;
1271}
1272
Willy Tarreau79e57332018-10-02 16:01:16 +02001273/* Accepts exactly 1 argument of type userlist */
1274static int smp_fetch_http_auth(const struct arg *args, struct sample *smp, const char *kw, void *private)
1275{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001276 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001277 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
Willy Tarreau79e57332018-10-02 16:01:16 +02001278
1279 if (!args || args->type != ARGT_USR)
1280 return 0;
1281
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001282 if (!htx)
1283 return 0;
1284 if (!get_http_auth(smp, htx))
1285 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001286
1287 smp->data.type = SMP_T_BOOL;
1288 smp->data.u.sint = check_user(args->data.usr, smp->strm->txn->auth.user,
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001289 smp->strm->txn->auth.pass);
Willy Tarreau79e57332018-10-02 16:01:16 +02001290 return 1;
1291}
1292
1293/* Accepts exactly 1 argument of type userlist */
1294static int smp_fetch_http_auth_grp(const struct arg *args, struct sample *smp, const char *kw, void *private)
1295{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001296 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001297 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
Christopher Faulet89dc4992019-04-17 12:02:59 +02001298
Willy Tarreau79e57332018-10-02 16:01:16 +02001299 if (!args || args->type != ARGT_USR)
1300 return 0;
1301
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001302 if (!htx)
1303 return 0;
1304 if (!get_http_auth(smp, htx))
1305 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001306
Willy Tarreau79e57332018-10-02 16:01:16 +02001307 /* if the user does not belong to the userlist or has a wrong password,
1308 * report that it unconditionally does not match. Otherwise we return
1309 * a string containing the username.
1310 */
1311 if (!check_user(args->data.usr, smp->strm->txn->auth.user,
1312 smp->strm->txn->auth.pass))
1313 return 0;
1314
1315 /* pat_match_auth() will need the user list */
1316 smp->ctx.a[0] = args->data.usr;
1317
1318 smp->data.type = SMP_T_STR;
1319 smp->flags = SMP_F_CONST;
1320 smp->data.u.str.area = smp->strm->txn->auth.user;
1321 smp->data.u.str.data = strlen(smp->strm->txn->auth.user);
1322
1323 return 1;
1324}
1325
1326/* Fetch a captured HTTP request header. The index is the position of
1327 * the "capture" option in the configuration file
1328 */
1329static int smp_fetch_capture_req_hdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
1330{
1331 struct proxy *fe = strm_fe(smp->strm);
1332 int idx;
1333
1334 if (!args || args->type != ARGT_SINT)
1335 return 0;
1336
1337 idx = args->data.sint;
1338
1339 if (idx > (fe->nb_req_cap - 1) || smp->strm->req_cap == NULL || smp->strm->req_cap[idx] == NULL)
1340 return 0;
1341
1342 smp->data.type = SMP_T_STR;
1343 smp->flags |= SMP_F_CONST;
1344 smp->data.u.str.area = smp->strm->req_cap[idx];
1345 smp->data.u.str.data = strlen(smp->strm->req_cap[idx]);
1346
1347 return 1;
1348}
1349
1350/* Fetch a captured HTTP response header. The index is the position of
1351 * the "capture" option in the configuration file
1352 */
1353static int smp_fetch_capture_res_hdr(const struct arg *args, struct sample *smp, const char *kw, void *private)
1354{
1355 struct proxy *fe = strm_fe(smp->strm);
1356 int idx;
1357
1358 if (!args || args->type != ARGT_SINT)
1359 return 0;
1360
1361 idx = args->data.sint;
1362
1363 if (idx > (fe->nb_rsp_cap - 1) || smp->strm->res_cap == NULL || smp->strm->res_cap[idx] == NULL)
1364 return 0;
1365
1366 smp->data.type = SMP_T_STR;
1367 smp->flags |= SMP_F_CONST;
1368 smp->data.u.str.area = smp->strm->res_cap[idx];
1369 smp->data.u.str.data = strlen(smp->strm->res_cap[idx]);
1370
1371 return 1;
1372}
1373
1374/* Extracts the METHOD in the HTTP request, the txn->uri should be filled before the call */
1375static int smp_fetch_capture_req_method(const struct arg *args, struct sample *smp, const char *kw, void *private)
1376{
1377 struct buffer *temp;
1378 struct http_txn *txn = smp->strm->txn;
1379 char *ptr;
1380
1381 if (!txn || !txn->uri)
1382 return 0;
1383
1384 ptr = txn->uri;
1385
1386 while (*ptr != ' ' && *ptr != '\0') /* find first space */
1387 ptr++;
1388
1389 temp = get_trash_chunk();
1390 temp->area = txn->uri;
1391 temp->data = ptr - txn->uri;
1392 smp->data.u.str = *temp;
1393 smp->data.type = SMP_T_STR;
1394 smp->flags = SMP_F_CONST;
1395
1396 return 1;
1397
1398}
1399
1400/* Extracts the path in the HTTP request, the txn->uri should be filled before the call */
1401static int smp_fetch_capture_req_uri(const struct arg *args, struct sample *smp, const char *kw, void *private)
1402{
1403 struct http_txn *txn = smp->strm->txn;
1404 struct ist path;
1405 const char *ptr;
1406
1407 if (!txn || !txn->uri)
1408 return 0;
1409
1410 ptr = txn->uri;
1411
1412 while (*ptr != ' ' && *ptr != '\0') /* find first space */
1413 ptr++;
1414
1415 if (!*ptr)
1416 return 0;
1417
Christopher Faulet78337bb2018-11-15 14:35:18 +01001418 /* skip the first space and find space after URI */
1419 path = ist2(++ptr, 0);
1420 while (*ptr != ' ' && *ptr != '\0')
1421 ptr++;
1422 path.len = ptr - path.ptr;
Willy Tarreau79e57332018-10-02 16:01:16 +02001423
Christopher Faulet78337bb2018-11-15 14:35:18 +01001424 path = http_get_path(path);
Tim Duesterhused526372020-03-05 17:56:33 +01001425 if (!isttest(path))
Willy Tarreau79e57332018-10-02 16:01:16 +02001426 return 0;
1427
1428 smp->data.u.str.area = path.ptr;
1429 smp->data.u.str.data = path.len;
1430 smp->data.type = SMP_T_STR;
1431 smp->flags = SMP_F_CONST;
1432
1433 return 1;
1434}
1435
1436/* Retrieves the HTTP version from the request (either 1.0 or 1.1) and emits it
1437 * as a string (either "HTTP/1.0" or "HTTP/1.1").
1438 */
1439static int smp_fetch_capture_req_ver(const struct arg *args, struct sample *smp, const char *kw, void *private)
1440{
1441 struct http_txn *txn = smp->strm->txn;
1442
Christopher Faulet711ed6a2019-07-16 14:16:10 +02001443 if (!txn || txn->req.msg_state >= HTTP_MSG_BODY)
Willy Tarreau79e57332018-10-02 16:01:16 +02001444 return 0;
1445
1446 if (txn->req.flags & HTTP_MSGF_VER_11)
1447 smp->data.u.str.area = "HTTP/1.1";
1448 else
1449 smp->data.u.str.area = "HTTP/1.0";
1450
1451 smp->data.u.str.data = 8;
1452 smp->data.type = SMP_T_STR;
1453 smp->flags = SMP_F_CONST;
1454 return 1;
1455
1456}
1457
1458/* Retrieves the HTTP version from the response (either 1.0 or 1.1) and emits it
1459 * as a string (either "HTTP/1.0" or "HTTP/1.1").
1460 */
1461static int smp_fetch_capture_res_ver(const struct arg *args, struct sample *smp, const char *kw, void *private)
1462{
1463 struct http_txn *txn = smp->strm->txn;
1464
Christopher Faulet711ed6a2019-07-16 14:16:10 +02001465 if (!txn || txn->rsp.msg_state >= HTTP_MSG_BODY)
Willy Tarreau79e57332018-10-02 16:01:16 +02001466 return 0;
1467
1468 if (txn->rsp.flags & HTTP_MSGF_VER_11)
1469 smp->data.u.str.area = "HTTP/1.1";
1470 else
1471 smp->data.u.str.area = "HTTP/1.0";
1472
1473 smp->data.u.str.data = 8;
1474 smp->data.type = SMP_T_STR;
1475 smp->flags = SMP_F_CONST;
1476 return 1;
1477
1478}
1479
1480/* Iterate over all cookies present in a message. The context is stored in
1481 * smp->ctx.a[0] for the in-header position, smp->ctx.a[1] for the
1482 * end-of-header-value, and smp->ctx.a[2] for the hdr_ctx. Depending on
1483 * the direction, multiple cookies may be parsed on the same line or not.
1484 * The cookie name is in args and the name length in args->data.str.len.
1485 * Accepts exactly 1 argument of type string. If the input options indicate
1486 * that no iterating is desired, then only last value is fetched if any.
1487 * The returned sample is of type CSTR. Can be used to parse cookies in other
1488 * files.
1489 */
1490static int smp_fetch_cookie(const struct arg *args, struct sample *smp, const char *kw, void *private)
1491{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001492 /* possible keywords: req.cookie / cookie / cook, res.cookie / scook / set-cookie */
1493 struct channel *chn = ((kw[0] == 'c' || kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001494 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1495 struct http_hdr_ctx *ctx = smp->ctx.a[2];
1496 struct ist hdr;
Willy Tarreau79e57332018-10-02 16:01:16 +02001497 int occ = 0;
1498 int found = 0;
1499
1500 if (!args || args->type != ARGT_STR)
1501 return 0;
1502
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001503 if (!ctx) {
1504 /* first call */
1505 ctx = &static_http_hdr_ctx;
1506 ctx->blk = NULL;
1507 smp->ctx.a[2] = ctx;
1508 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001509
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001510 if (!htx)
1511 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001512
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001513 hdr = (!(chn->flags & CF_ISRESP) ? ist("Cookie") : ist("Set-Cookie"));
Willy Tarreau79e57332018-10-02 16:01:16 +02001514
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001515 if (!occ && !(smp->opt & SMP_OPT_ITERATE))
1516 /* no explicit occurrence and single fetch => last cookie by default */
1517 occ = -1;
Willy Tarreau79e57332018-10-02 16:01:16 +02001518
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001519 /* OK so basically here, either we want only one value and it's the
1520 * last one, or we want to iterate over all of them and we fetch the
1521 * next one.
1522 */
Willy Tarreau79e57332018-10-02 16:01:16 +02001523
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001524 if (!(smp->flags & SMP_F_NOT_LAST)) {
1525 /* search for the header from the beginning, we must first initialize
1526 * the search parameters.
Willy Tarreau79e57332018-10-02 16:01:16 +02001527 */
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001528 smp->ctx.a[0] = NULL;
1529 ctx->blk = NULL;
Willy Tarreau79e57332018-10-02 16:01:16 +02001530 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001531
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001532 smp->flags |= SMP_F_VOL_HDR;
1533 while (1) {
1534 /* Note: smp->ctx.a[0] == NULL every time we need to fetch a new header */
1535 if (!smp->ctx.a[0]) {
1536 if (!http_find_header(htx, hdr, ctx, 0))
1537 goto out;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001538
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001539 if (ctx->value.len < args->data.str.data + 1)
1540 continue;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001541
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001542 smp->ctx.a[0] = ctx->value.ptr;
1543 smp->ctx.a[1] = smp->ctx.a[0] + ctx->value.len;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001544 }
1545
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001546 smp->data.type = SMP_T_STR;
1547 smp->flags |= SMP_F_CONST;
1548 smp->ctx.a[0] = http_extract_cookie_value(smp->ctx.a[0], smp->ctx.a[1],
1549 args->data.str.area, args->data.str.data,
1550 (smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ,
1551 &smp->data.u.str.area,
1552 &smp->data.u.str.data);
1553 if (smp->ctx.a[0]) {
1554 found = 1;
1555 if (occ >= 0) {
1556 /* one value was returned into smp->data.u.str.{str,len} */
1557 smp->flags |= SMP_F_NOT_LAST;
1558 return 1;
Willy Tarreau79e57332018-10-02 16:01:16 +02001559 }
1560 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001561 /* if we're looking for last occurrence, let's loop */
Willy Tarreau79e57332018-10-02 16:01:16 +02001562 }
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001563
Willy Tarreau79e57332018-10-02 16:01:16 +02001564 /* all cookie headers and values were scanned. If we're looking for the
1565 * last occurrence, we may return it now.
1566 */
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001567 out:
Willy Tarreau79e57332018-10-02 16:01:16 +02001568 smp->flags &= ~SMP_F_NOT_LAST;
1569 return found;
1570}
1571
Christopher Fauletc1f40dd2019-05-16 10:07:30 +02001572/* Same than smp_fetch_cookie() but only relies on the sample direction to
1573 * choose the right channel. So instead of duplicating the code, we just change
1574 * the keyword and then fallback on smp_fetch_cookie().
1575 */
1576static int smp_fetch_chn_cookie(const struct arg *args, struct sample *smp, const char *kw, void *private)
1577{
1578 kw = ((smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ ? "req.cook" : "res.cook");
1579 return smp_fetch_cookie(args, smp, kw, private);
1580}
1581
Willy Tarreau79e57332018-10-02 16:01:16 +02001582/* Iterate over all cookies present in a request to count how many occurrences
1583 * match the name in args and args->data.str.len. If <multi> is non-null, then
1584 * multiple cookies may be parsed on the same line. The returned sample is of
1585 * type UINT. Accepts exactly 1 argument of type string.
1586 */
1587static int smp_fetch_cookie_cnt(const struct arg *args, struct sample *smp, const char *kw, void *private)
1588{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001589 /* possible keywords: req.cook_cnt / cook_cnt, res.cook_cnt / scook_cnt */
1590 struct channel *chn = ((kw[0] == 'c' || kw[2] == 'q') ? SMP_REQ_CHN(smp) : SMP_RES_CHN(smp));
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001591 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1592 struct http_hdr_ctx ctx;
1593 struct ist hdr;
Willy Tarreau79e57332018-10-02 16:01:16 +02001594 char *val_beg, *val_end;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001595 int cnt;
Willy Tarreau79e57332018-10-02 16:01:16 +02001596
1597 if (!args || args->type != ARGT_STR)
1598 return 0;
1599
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001600 if (!htx)
1601 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001602
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001603 hdr = (!(chn->flags & CF_ISRESP) ? ist("Cookie") : ist("Set-Cookie"));
Willy Tarreau79e57332018-10-02 16:01:16 +02001604
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001605 val_end = val_beg = NULL;
1606 ctx.blk = NULL;
1607 cnt = 0;
1608 while (1) {
1609 /* Note: val_beg == NULL every time we need to fetch a new header */
1610 if (!val_beg) {
1611 if (!http_find_header(htx, hdr, &ctx, 0))
1612 break;
Willy Tarreau79e57332018-10-02 16:01:16 +02001613
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001614 if (ctx.value.len < args->data.str.data + 1)
1615 continue;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001616
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001617 val_beg = ctx.value.ptr;
1618 val_end = val_beg + ctx.value.len;
Willy Tarreau79e57332018-10-02 16:01:16 +02001619 }
1620
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001621 smp->data.type = SMP_T_STR;
1622 smp->flags |= SMP_F_CONST;
1623 while ((val_beg = http_extract_cookie_value(val_beg, val_end,
1624 args->data.str.area, args->data.str.data,
1625 (smp->opt & SMP_OPT_DIR) == SMP_OPT_DIR_REQ,
1626 &smp->data.u.str.area,
1627 &smp->data.u.str.data))) {
1628 cnt++;
Willy Tarreau79e57332018-10-02 16:01:16 +02001629 }
1630 }
1631
1632 smp->data.type = SMP_T_SINT;
1633 smp->data.u.sint = cnt;
1634 smp->flags |= SMP_F_VOL_HDR;
1635 return 1;
1636}
1637
1638/* Fetch an cookie's integer value. The integer value is returned. It
1639 * takes a mandatory argument of type string. It relies on smp_fetch_cookie().
1640 */
1641static int smp_fetch_cookie_val(const struct arg *args, struct sample *smp, const char *kw, void *private)
1642{
1643 int ret = smp_fetch_cookie(args, smp, kw, private);
1644
1645 if (ret > 0) {
1646 smp->data.type = SMP_T_SINT;
1647 smp->data.u.sint = strl2ic(smp->data.u.str.area,
1648 smp->data.u.str.data);
1649 }
1650
1651 return ret;
1652}
1653
1654/************************************************************************/
1655/* The code below is dedicated to sample fetches */
1656/************************************************************************/
1657
1658/* This scans a URL-encoded query string. It takes an optionally wrapping
1659 * string whose first contigous chunk has its beginning in ctx->a[0] and end
1660 * in ctx->a[1], and the optional second part in (ctx->a[2]..ctx->a[3]). The
1661 * pointers are updated for next iteration before leaving.
1662 */
1663static int smp_fetch_param(char delim, const char *name, int name_len, const struct arg *args, struct sample *smp, const char *kw, void *private)
1664{
1665 const char *vstart, *vend;
1666 struct buffer *temp;
1667 const char **chunks = (const char **)smp->ctx.a;
1668
1669 if (!http_find_next_url_param(chunks, name, name_len,
1670 &vstart, &vend, delim))
1671 return 0;
1672
1673 /* Create sample. If the value is contiguous, return the pointer as CONST,
1674 * if the value is wrapped, copy-it in a buffer.
1675 */
1676 smp->data.type = SMP_T_STR;
1677 if (chunks[2] &&
1678 vstart >= chunks[0] && vstart <= chunks[1] &&
1679 vend >= chunks[2] && vend <= chunks[3]) {
1680 /* Wrapped case. */
1681 temp = get_trash_chunk();
1682 memcpy(temp->area, vstart, chunks[1] - vstart);
1683 memcpy(temp->area + ( chunks[1] - vstart ), chunks[2],
1684 vend - chunks[2]);
1685 smp->data.u.str.area = temp->area;
1686 smp->data.u.str.data = ( chunks[1] - vstart ) + ( vend - chunks[2] );
1687 } else {
1688 /* Contiguous case. */
1689 smp->data.u.str.area = (char *)vstart;
1690 smp->data.u.str.data = vend - vstart;
1691 smp->flags = SMP_F_VOL_1ST | SMP_F_CONST;
1692 }
1693
1694 /* Update context, check wrapping. */
1695 chunks[0] = vend;
1696 if (chunks[2] && vend >= chunks[2] && vend <= chunks[3]) {
1697 chunks[1] = chunks[3];
1698 chunks[2] = NULL;
1699 }
1700
1701 if (chunks[0] < chunks[1])
1702 smp->flags |= SMP_F_NOT_LAST;
1703
1704 return 1;
1705}
1706
1707/* This function iterates over each parameter of the query string. It uses
1708 * ctx->a[0] and ctx->a[1] to store the beginning and end of the current
1709 * parameter. Since it uses smp_fetch_param(), ctx->a[2..3] are both NULL.
1710 * An optional parameter name is passed in args[0], otherwise any parameter is
1711 * considered. It supports an optional delimiter argument for the beginning of
1712 * the string in args[1], which defaults to "?".
1713 */
1714static int smp_fetch_url_param(const struct arg *args, struct sample *smp, const char *kw, void *private)
1715{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001716 struct channel *chn = SMP_REQ_CHN(smp);
Willy Tarreau79e57332018-10-02 16:01:16 +02001717 char delim = '?';
1718 const char *name;
1719 int name_len;
1720
1721 if (!args ||
1722 (args[0].type && args[0].type != ARGT_STR) ||
1723 (args[1].type && args[1].type != ARGT_STR))
1724 return 0;
1725
1726 name = "";
1727 name_len = 0;
1728 if (args->type == ARGT_STR) {
1729 name = args->data.str.area;
1730 name_len = args->data.str.data;
1731 }
1732
1733 if (args[1].type)
1734 delim = *args[1].data.str.area;
1735
1736 if (!smp->ctx.a[0]) { // first call, find the query string
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001737 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1738 struct htx_sl *sl;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001739
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001740 if (!htx)
1741 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001742
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001743 sl = http_get_stline(htx);
1744 smp->ctx.a[0] = http_find_param_list(HTX_SL_REQ_UPTR(sl), HTX_SL_REQ_ULEN(sl), delim);
1745 if (!smp->ctx.a[0])
1746 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001747
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001748 smp->ctx.a[1] = HTX_SL_REQ_UPTR(sl) + HTX_SL_REQ_ULEN(sl);
Willy Tarreau79e57332018-10-02 16:01:16 +02001749
1750 /* Assume that the context is filled with NULL pointer
1751 * before the first call.
1752 * smp->ctx.a[2] = NULL;
1753 * smp->ctx.a[3] = NULL;
1754 */
1755 }
1756
1757 return smp_fetch_param(delim, name, name_len, args, smp, kw, private);
1758}
1759
1760/* This function iterates over each parameter of the body. This requires
1761 * that the body has been waited for using http-buffer-request. It uses
1762 * ctx->a[0] and ctx->a[1] to store the beginning and end of the first
1763 * contigous part of the body, and optionally ctx->a[2..3] to reference the
1764 * optional second part if the body wraps at the end of the buffer. An optional
1765 * parameter name is passed in args[0], otherwise any parameter is considered.
1766 */
1767static int smp_fetch_body_param(const struct arg *args, struct sample *smp, const char *kw, void *private)
1768{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001769 struct channel *chn = SMP_REQ_CHN(smp);
Willy Tarreau79e57332018-10-02 16:01:16 +02001770 const char *name;
1771 int name_len;
1772
1773 if (!args || (args[0].type && args[0].type != ARGT_STR))
1774 return 0;
1775
1776 name = "";
1777 name_len = 0;
1778 if (args[0].type == ARGT_STR) {
1779 name = args[0].data.str.area;
1780 name_len = args[0].data.str.data;
1781 }
1782
1783 if (!smp->ctx.a[0]) { // first call, find the query string
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001784 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1785 struct buffer *temp;
1786 int32_t pos;
Willy Tarreau79e57332018-10-02 16:01:16 +02001787
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001788 if (!htx)
1789 return 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001790
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001791 temp = get_trash_chunk();
1792 for (pos = htx_get_first(htx); pos != -1; pos = htx_get_next(htx, pos)) {
1793 struct htx_blk *blk = htx_get_blk(htx, pos);
1794 enum htx_blk_type type = htx_get_blk_type(blk);
Willy Tarreau79e57332018-10-02 16:01:16 +02001795
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001796 if (type == HTX_BLK_EOM || type == HTX_BLK_TLR || type == HTX_BLK_EOT)
1797 break;
1798 if (type == HTX_BLK_DATA) {
Christopher Faulet53a899b2019-10-08 16:38:42 +02001799 if (!h1_format_htx_data(htx_get_blk_value(htx, blk), temp, 0))
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001800 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001801 }
Willy Tarreau79e57332018-10-02 16:01:16 +02001802 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001803
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001804 smp->ctx.a[0] = temp->area;
1805 smp->ctx.a[1] = temp->area + temp->data;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001806
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001807 /* Assume that the context is filled with NULL pointer
1808 * before the first call.
1809 * smp->ctx.a[2] = NULL;
1810 * smp->ctx.a[3] = NULL;
1811 */
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001812
Willy Tarreau79e57332018-10-02 16:01:16 +02001813 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001814
Willy Tarreau79e57332018-10-02 16:01:16 +02001815 return smp_fetch_param('&', name, name_len, args, smp, kw, private);
1816}
1817
1818/* Return the signed integer value for the specified url parameter (see url_param
1819 * above).
1820 */
1821static int smp_fetch_url_param_val(const struct arg *args, struct sample *smp, const char *kw, void *private)
1822{
1823 int ret = smp_fetch_url_param(args, smp, kw, private);
1824
1825 if (ret > 0) {
1826 smp->data.type = SMP_T_SINT;
1827 smp->data.u.sint = strl2ic(smp->data.u.str.area,
1828 smp->data.u.str.data);
1829 }
1830
1831 return ret;
1832}
1833
1834/* This produces a 32-bit hash of the concatenation of the first occurrence of
1835 * the Host header followed by the path component if it begins with a slash ('/').
1836 * This means that '*' will not be added, resulting in exactly the first Host
1837 * entry. If no Host header is found, then the path is used. The resulting value
1838 * is hashed using the url hash followed by a full avalanche hash and provides a
1839 * 32-bit integer value. This fetch is useful for tracking per-URL activity on
1840 * high-traffic sites without having to store whole paths.
1841 * this differs from the base32 functions in that it includes the url parameters
1842 * as well as the path
1843 */
1844static int smp_fetch_url32(const struct arg *args, struct sample *smp, const char *kw, void *private)
1845{
Christopher Faulet89dc4992019-04-17 12:02:59 +02001846 struct channel *chn = SMP_REQ_CHN(smp);
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001847 struct htx *htx = smp_prefetch_htx(smp, chn, 1);
1848 struct http_hdr_ctx ctx;
1849 struct htx_sl *sl;
1850 struct ist path;
Willy Tarreau79e57332018-10-02 16:01:16 +02001851 unsigned int hash = 0;
Willy Tarreau79e57332018-10-02 16:01:16 +02001852
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001853 if (!htx)
1854 return 0;
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001855
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001856 ctx.blk = NULL;
1857 if (http_find_header(htx, ist("Host"), &ctx, 1)) {
1858 /* OK we have the header value in ctx.value */
1859 while (ctx.value.len--)
1860 hash = *(ctx.value.ptr++) + (hash << 6) + (hash << 16) - hash;
Willy Tarreau79e57332018-10-02 16:01:16 +02001861 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001862
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001863 /* now retrieve the path */
1864 sl = http_get_stline(htx);
1865 path = http_get_path(htx_sl_req_uri(sl));
Christopher Faulet6d1dd462019-07-15 14:36:03 +02001866 if (path.len && *(path.ptr) == '/') {
1867 while (path.len--)
1868 hash = *(path.ptr++) + (hash << 6) + (hash << 16) - hash;
Willy Tarreau79e57332018-10-02 16:01:16 +02001869 }
Christopher Faulet311c7ea2018-10-24 21:41:55 +02001870
Willy Tarreau79e57332018-10-02 16:01:16 +02001871 hash = full_hash(hash);
1872
1873 smp->data.type = SMP_T_SINT;
1874 smp->data.u.sint = hash;
1875 smp->flags = SMP_F_VOL_1ST;
1876 return 1;
1877}
1878
1879/* This concatenates the source address with the 32-bit hash of the Host and
1880 * URL as returned by smp_fetch_base32(). The idea is to have per-source and
1881 * per-url counters. The result is a binary block from 8 to 20 bytes depending
1882 * on the source address length. The URL hash is stored before the address so
1883 * that in environments where IPv6 is insignificant, truncating the output to
1884 * 8 bytes would still work.
1885 */
1886static int smp_fetch_url32_src(const struct arg *args, struct sample *smp, const char *kw, void *private)
1887{
1888 struct buffer *temp;
1889 struct connection *cli_conn = objt_conn(smp->sess->origin);
1890
Willy Tarreaucd7ca792019-07-17 16:57:03 +02001891 if (!cli_conn || !conn_get_src(cli_conn))
Willy Tarreau79e57332018-10-02 16:01:16 +02001892 return 0;
1893
1894 if (!smp_fetch_url32(args, smp, kw, private))
1895 return 0;
1896
1897 temp = get_trash_chunk();
1898 *(unsigned int *) temp->area = htonl(smp->data.u.sint);
1899 temp->data += sizeof(unsigned int);
1900
Willy Tarreau9a1efe12019-07-17 17:13:50 +02001901 switch (cli_conn->src->ss_family) {
Willy Tarreau79e57332018-10-02 16:01:16 +02001902 case AF_INET:
1903 memcpy(temp->area + temp->data,
Willy Tarreau9a1efe12019-07-17 17:13:50 +02001904 &((struct sockaddr_in *)cli_conn->src)->sin_addr,
Willy Tarreau79e57332018-10-02 16:01:16 +02001905 4);
1906 temp->data += 4;
1907 break;
1908 case AF_INET6:
1909 memcpy(temp->area + temp->data,
Willy Tarreau9a1efe12019-07-17 17:13:50 +02001910 &((struct sockaddr_in6 *)cli_conn->src)->sin6_addr,
Willy Tarreau79e57332018-10-02 16:01:16 +02001911 16);
1912 temp->data += 16;
1913 break;
1914 default:
1915 return 0;
1916 }
1917
1918 smp->data.u.str = *temp;
1919 smp->data.type = SMP_T_BIN;
1920 return 1;
1921}
1922
1923/************************************************************************/
1924/* Other utility functions */
1925/************************************************************************/
1926
1927/* This function is used to validate the arguments passed to any "hdr" fetch
1928 * keyword. These keywords support an optional positive or negative occurrence
1929 * number. We must ensure that the number is greater than -MAX_HDR_HISTORY. It
1930 * is assumed that the types are already the correct ones. Returns 0 on error,
1931 * non-zero if OK. If <err> is not NULL, it will be filled with a pointer to an
1932 * error message in case of error, that the caller is responsible for freeing.
1933 * The initial location must either be freeable or NULL.
1934 * Note: this function's pointer is checked from Lua.
1935 */
1936int val_hdr(struct arg *arg, char **err_msg)
1937{
1938 if (arg && arg[1].type == ARGT_SINT && arg[1].data.sint < -MAX_HDR_HISTORY) {
1939 memprintf(err_msg, "header occurrence must be >= %d", -MAX_HDR_HISTORY);
1940 return 0;
1941 }
1942 return 1;
1943}
1944
1945/************************************************************************/
1946/* All supported sample fetch keywords must be declared here. */
1947/************************************************************************/
1948
1949/* Note: must not be declared <const> as its list will be overwritten */
1950static struct sample_fetch_kw_list sample_fetch_keywords = {ILH, {
1951 { "base", smp_fetch_base, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
1952 { "base32", smp_fetch_base32, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
1953 { "base32+src", smp_fetch_base32_src, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
1954
1955 /* capture are allocated and are permanent in the stream */
1956 { "capture.req.hdr", smp_fetch_capture_req_hdr, ARG1(1,SINT), NULL, SMP_T_STR, SMP_USE_HRQHP },
1957
1958 /* retrieve these captures from the HTTP logs */
1959 { "capture.req.method", smp_fetch_capture_req_method, 0, NULL, SMP_T_STR, SMP_USE_HRQHP },
1960 { "capture.req.uri", smp_fetch_capture_req_uri, 0, NULL, SMP_T_STR, SMP_USE_HRQHP },
1961 { "capture.req.ver", smp_fetch_capture_req_ver, 0, NULL, SMP_T_STR, SMP_USE_HRQHP },
1962
1963 { "capture.res.hdr", smp_fetch_capture_res_hdr, ARG1(1,SINT), NULL, SMP_T_STR, SMP_USE_HRSHP },
1964 { "capture.res.ver", smp_fetch_capture_res_ver, 0, NULL, SMP_T_STR, SMP_USE_HRQHP },
1965
1966 /* cookie is valid in both directions (eg: for "stick ...") but cook*
1967 * are only here to match the ACL's name, are request-only and are used
1968 * for ACL compatibility only.
1969 */
1970 { "cook", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
Christopher Fauletc1f40dd2019-05-16 10:07:30 +02001971 { "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 +02001972 { "cook_cnt", smp_fetch_cookie_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
1973 { "cook_val", smp_fetch_cookie_val, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
1974
1975 /* hdr is valid in both directions (eg: for "stick ...") but hdr_* are
1976 * only here to match the ACL's name, are request-only and are used for
1977 * ACL compatibility only.
1978 */
Christopher Fauletc1f40dd2019-05-16 10:07:30 +02001979 { "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 +02001980 { "hdr_cnt", smp_fetch_hdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
1981 { "hdr_ip", smp_fetch_hdr_ip, ARG2(0,STR,SINT), val_hdr, SMP_T_IPV4, SMP_USE_HRQHV },
1982 { "hdr_val", smp_fetch_hdr_val, ARG2(0,STR,SINT), val_hdr, SMP_T_SINT, SMP_USE_HRQHV },
1983
Christopher Fauleta4063562019-08-02 11:51:37 +02001984 { "http_auth_type", smp_fetch_http_auth_type, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
1985 { "http_auth_user", smp_fetch_http_auth_user, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
1986 { "http_auth_pass", smp_fetch_http_auth_pass, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
Willy Tarreau79e57332018-10-02 16:01:16 +02001987 { "http_auth", smp_fetch_http_auth, ARG1(1,USR), NULL, SMP_T_BOOL, SMP_USE_HRQHV },
1988 { "http_auth_group", smp_fetch_http_auth_grp, ARG1(1,USR), NULL, SMP_T_STR, SMP_USE_HRQHV },
1989 { "http_first_req", smp_fetch_http_first_req, 0, NULL, SMP_T_BOOL, SMP_USE_HRQHP },
1990 { "method", smp_fetch_meth, 0, NULL, SMP_T_METH, SMP_USE_HRQHP },
1991 { "path", smp_fetch_path, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
1992 { "query", smp_fetch_query, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
1993
1994 /* HTTP protocol on the request path */
1995 { "req.proto_http", smp_fetch_proto_http, 0, NULL, SMP_T_BOOL, SMP_USE_HRQHP },
1996 { "req_proto_http", smp_fetch_proto_http, 0, NULL, SMP_T_BOOL, SMP_USE_HRQHP },
1997
1998 /* HTTP version on the request path */
1999 { "req.ver", smp_fetch_rqver, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2000 { "req_ver", smp_fetch_rqver, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2001
2002 { "req.body", smp_fetch_body, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
2003 { "req.body_len", smp_fetch_body_len, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2004 { "req.body_size", smp_fetch_body_size, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2005 { "req.body_param", smp_fetch_body_param, ARG1(0,STR), NULL, SMP_T_BIN, SMP_USE_HRQHV },
2006
2007 { "req.hdrs", smp_fetch_hdrs, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
2008 { "req.hdrs_bin", smp_fetch_hdrs_bin, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
2009
2010 /* HTTP version on the response path */
2011 { "res.ver", smp_fetch_stver, 0, NULL, SMP_T_STR, SMP_USE_HRSHV },
2012 { "resp_ver", smp_fetch_stver, 0, NULL, SMP_T_STR, SMP_USE_HRSHV },
2013
2014 /* explicit req.{cook,hdr} are used to force the fetch direction to be request-only */
2015 { "req.cook", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2016 { "req.cook_cnt", smp_fetch_cookie_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2017 { "req.cook_val", smp_fetch_cookie_val, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2018
2019 { "req.fhdr", smp_fetch_fhdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRQHV },
2020 { "req.fhdr_cnt", smp_fetch_fhdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2021 { "req.hdr", smp_fetch_hdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRQHV },
2022 { "req.hdr_cnt", smp_fetch_hdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2023 { "req.hdr_ip", smp_fetch_hdr_ip, ARG2(0,STR,SINT), val_hdr, SMP_T_IPV4, SMP_USE_HRQHV },
2024 { "req.hdr_names", smp_fetch_hdr_names, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2025 { "req.hdr_val", smp_fetch_hdr_val, ARG2(0,STR,SINT), val_hdr, SMP_T_SINT, SMP_USE_HRQHV },
2026
2027 /* explicit req.{cook,hdr} are used to force the fetch direction to be response-only */
2028 { "res.cook", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRSHV },
2029 { "res.cook_cnt", smp_fetch_cookie_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2030 { "res.cook_val", smp_fetch_cookie_val, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2031
2032 { "res.fhdr", smp_fetch_fhdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRSHV },
2033 { "res.fhdr_cnt", smp_fetch_fhdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2034 { "res.hdr", smp_fetch_hdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRSHV },
2035 { "res.hdr_cnt", smp_fetch_hdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2036 { "res.hdr_ip", smp_fetch_hdr_ip, ARG2(0,STR,SINT), val_hdr, SMP_T_IPV4, SMP_USE_HRSHV },
2037 { "res.hdr_names", smp_fetch_hdr_names, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRSHV },
2038 { "res.hdr_val", smp_fetch_hdr_val, ARG2(0,STR,SINT), val_hdr, SMP_T_SINT, SMP_USE_HRSHV },
2039
2040 /* scook is valid only on the response and is used for ACL compatibility */
2041 { "scook", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRSHV },
2042 { "scook_cnt", smp_fetch_cookie_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2043 { "scook_val", smp_fetch_cookie_val, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2044 { "set-cookie", smp_fetch_cookie, ARG1(0,STR), NULL, SMP_T_STR, SMP_USE_HRSHV }, /* deprecated */
2045
2046 /* shdr is valid only on the response and is used for ACL compatibility */
2047 { "shdr", smp_fetch_hdr, ARG2(0,STR,SINT), val_hdr, SMP_T_STR, SMP_USE_HRSHV },
2048 { "shdr_cnt", smp_fetch_hdr_cnt, ARG1(0,STR), NULL, SMP_T_SINT, SMP_USE_HRSHV },
2049 { "shdr_ip", smp_fetch_hdr_ip, ARG2(0,STR,SINT), val_hdr, SMP_T_IPV4, SMP_USE_HRSHV },
2050 { "shdr_val", smp_fetch_hdr_val, ARG2(0,STR,SINT), val_hdr, SMP_T_SINT, SMP_USE_HRSHV },
2051
2052 { "status", smp_fetch_stcode, 0, NULL, SMP_T_SINT, SMP_USE_HRSHP },
2053 { "unique-id", smp_fetch_uniqueid, 0, NULL, SMP_T_STR, SMP_SRC_L4SRV },
2054 { "url", smp_fetch_url, 0, NULL, SMP_T_STR, SMP_USE_HRQHV },
2055 { "url32", smp_fetch_url32, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2056 { "url32+src", smp_fetch_url32_src, 0, NULL, SMP_T_BIN, SMP_USE_HRQHV },
2057 { "url_ip", smp_fetch_url_ip, 0, NULL, SMP_T_IPV4, SMP_USE_HRQHV },
2058 { "url_port", smp_fetch_url_port, 0, NULL, SMP_T_SINT, SMP_USE_HRQHV },
2059 { "url_param", smp_fetch_url_param, ARG2(0,STR,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2060 { "urlp" , smp_fetch_url_param, ARG2(0,STR,STR), NULL, SMP_T_STR, SMP_USE_HRQHV },
2061 { "urlp_val", smp_fetch_url_param_val, ARG2(0,STR,STR), NULL, SMP_T_SINT, SMP_USE_HRQHV },
2062 { /* END */ },
2063}};
2064
Willy Tarreau0108d902018-11-25 19:14:37 +01002065INITCALL1(STG_REGISTER, sample_register_fetches, &sample_fetch_keywords);
Willy Tarreau79e57332018-10-02 16:01:16 +02002066
2067/*
2068 * Local variables:
2069 * c-indent-level: 8
2070 * c-basic-offset: 8
2071 * End:
2072 */