blob: 4914d51d1300fd6c0bd6d66fc95278e2d5bb69d7 [file] [log] [blame]
Willy Tarreau0da5b3b2017-09-21 09:30:46 +02001/*
2 * include/proto/h1.h
3 * This file contains HTTP/1 protocol definitions.
4 *
5 * Copyright (C) 2000-2017 Willy Tarreau - w@1wt.eu
6 *
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation, version 2.1
10 * exclusively.
11 *
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with this library; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22#ifndef _PROTO_H1_H
23#define _PROTO_H1_H
24
Willy Tarreaudb4893d2017-09-21 08:40:02 +020025#include <common/buffer.h>
Willy Tarreau0da5b3b2017-09-21 09:30:46 +020026#include <common/compiler.h>
27#include <common/config.h>
Willy Tarreau35b51c62018-09-10 15:38:55 +020028#include <common/http.h>
Willy Tarreau794f9af2017-07-26 09:07:47 +020029#include <common/http-hdr.h>
Willy Tarreaudb4893d2017-09-21 08:40:02 +020030#include <common/standard.h>
Willy Tarreau0da5b3b2017-09-21 09:30:46 +020031#include <types/h1.h>
Willy Tarreau8740c8b2017-09-21 10:22:25 +020032#include <proto/hdr_idx.h>
Willy Tarreau0da5b3b2017-09-21 09:30:46 +020033
Willy Tarreau8740c8b2017-09-21 10:22:25 +020034const char *http_parse_reqline(struct http_msg *msg,
35 enum h1_state state, const char *ptr, const char *end,
36 unsigned int *ret_ptr, enum h1_state *ret_state);
37const char *http_parse_stsline(struct http_msg *msg,
38 enum h1_state state, const char *ptr, const char *end,
39 unsigned int *ret_ptr, enum h1_state *ret_state);
40void http_msg_analyzer(struct http_msg *msg, struct hdr_idx *idx);
Willy Tarreaudb4893d2017-09-21 08:40:02 +020041int http_forward_trailers(struct http_msg *msg);
Willy Tarreau794f9af2017-07-26 09:07:47 +020042int h1_headers_to_hdr_list(char *start, const char *stop,
43 struct http_hdr *hdr, unsigned int hdr_num,
Willy Tarreaua41393f2018-09-11 15:34:50 +020044 struct h1m *h1m, union h1_sl *slp);
Willy Tarreauf40e6822018-06-14 16:52:02 +020045int h1_measure_trailers(const struct buffer *buf, unsigned int ofs, unsigned int max);
Willy Tarreau0da5b3b2017-09-21 09:30:46 +020046
Willy Tarreau0da5b3b2017-09-21 09:30:46 +020047/* Macros used in the HTTP/1 parser, to check for the expected presence of
48 * certain bytes (ef: LF) or to skip to next byte and yield in case of failure.
49 */
50
51
52/* Expects to find an LF at <ptr>. If not, set <state> to <where> and jump to
53 * <bad>.
54 */
55#define EXPECT_LF_HERE(ptr, bad, state, where) \
56 do { \
57 if (unlikely(*(ptr) != '\n')) { \
58 state = (where); \
59 goto bad; \
60 } \
61 } while (0)
62
63/* Increments pointer <ptr>, continues to label <more> if it's still below
64 * pointer <end>, or goes to <stop> and sets <state> to <where> if the end
65 * of buffer was reached.
66 */
67#define EAT_AND_JUMP_OR_RETURN(ptr, end, more, stop, state, where) \
68 do { \
69 if (likely(++(ptr) < (end))) \
70 goto more; \
71 else { \
72 state = (where); \
73 goto stop; \
74 } \
75 } while (0)
76
Willy Tarreau801250e2018-09-11 11:45:04 +020077/* for debugging, reports the HTTP/1 message state name (legacy version) */
Willy Tarreau0da5b3b2017-09-21 09:30:46 +020078static inline const char *h1_msg_state_str(enum h1_state msg_state)
79{
80 switch (msg_state) {
81 case HTTP_MSG_RQBEFORE: return "MSG_RQBEFORE";
82 case HTTP_MSG_RQBEFORE_CR: return "MSG_RQBEFORE_CR";
83 case HTTP_MSG_RQMETH: return "MSG_RQMETH";
84 case HTTP_MSG_RQMETH_SP: return "MSG_RQMETH_SP";
85 case HTTP_MSG_RQURI: return "MSG_RQURI";
86 case HTTP_MSG_RQURI_SP: return "MSG_RQURI_SP";
87 case HTTP_MSG_RQVER: return "MSG_RQVER";
88 case HTTP_MSG_RQLINE_END: return "MSG_RQLINE_END";
89 case HTTP_MSG_RPBEFORE: return "MSG_RPBEFORE";
90 case HTTP_MSG_RPBEFORE_CR: return "MSG_RPBEFORE_CR";
91 case HTTP_MSG_RPVER: return "MSG_RPVER";
92 case HTTP_MSG_RPVER_SP: return "MSG_RPVER_SP";
93 case HTTP_MSG_RPCODE: return "MSG_RPCODE";
94 case HTTP_MSG_RPCODE_SP: return "MSG_RPCODE_SP";
95 case HTTP_MSG_RPREASON: return "MSG_RPREASON";
96 case HTTP_MSG_RPLINE_END: return "MSG_RPLINE_END";
97 case HTTP_MSG_HDR_FIRST: return "MSG_HDR_FIRST";
98 case HTTP_MSG_HDR_NAME: return "MSG_HDR_NAME";
99 case HTTP_MSG_HDR_COL: return "MSG_HDR_COL";
100 case HTTP_MSG_HDR_L1_SP: return "MSG_HDR_L1_SP";
101 case HTTP_MSG_HDR_L1_LF: return "MSG_HDR_L1_LF";
102 case HTTP_MSG_HDR_L1_LWS: return "MSG_HDR_L1_LWS";
103 case HTTP_MSG_HDR_VAL: return "MSG_HDR_VAL";
104 case HTTP_MSG_HDR_L2_LF: return "MSG_HDR_L2_LF";
105 case HTTP_MSG_HDR_L2_LWS: return "MSG_HDR_L2_LWS";
106 case HTTP_MSG_LAST_LF: return "MSG_LAST_LF";
107 case HTTP_MSG_ERROR: return "MSG_ERROR";
108 case HTTP_MSG_BODY: return "MSG_BODY";
109 case HTTP_MSG_100_SENT: return "MSG_100_SENT";
110 case HTTP_MSG_CHUNK_SIZE: return "MSG_CHUNK_SIZE";
111 case HTTP_MSG_DATA: return "MSG_DATA";
112 case HTTP_MSG_CHUNK_CRLF: return "MSG_CHUNK_CRLF";
113 case HTTP_MSG_TRAILERS: return "MSG_TRAILERS";
114 case HTTP_MSG_ENDING: return "MSG_ENDING";
115 case HTTP_MSG_DONE: return "MSG_DONE";
116 case HTTP_MSG_CLOSING: return "MSG_CLOSING";
117 case HTTP_MSG_CLOSED: return "MSG_CLOSED";
118 case HTTP_MSG_TUNNEL: return "MSG_TUNNEL";
119 default: return "MSG_??????";
120 }
121}
122
Willy Tarreau801250e2018-09-11 11:45:04 +0200123/* for debugging, reports the HTTP/1 message state name */
Dirkjan Bussinkc26c72d2018-09-14 14:30:25 +0200124static inline const char *h1m_state_str(enum h1m_state msg_state)
Willy Tarreau801250e2018-09-11 11:45:04 +0200125{
126 switch (msg_state) {
127 case H1_MSG_RQBEFORE: return "MSG_RQBEFORE";
128 case H1_MSG_RQBEFORE_CR: return "MSG_RQBEFORE_CR";
129 case H1_MSG_RQMETH: return "MSG_RQMETH";
130 case H1_MSG_RQMETH_SP: return "MSG_RQMETH_SP";
131 case H1_MSG_RQURI: return "MSG_RQURI";
132 case H1_MSG_RQURI_SP: return "MSG_RQURI_SP";
133 case H1_MSG_RQVER: return "MSG_RQVER";
134 case H1_MSG_RQLINE_END: return "MSG_RQLINE_END";
135 case H1_MSG_RPBEFORE: return "MSG_RPBEFORE";
136 case H1_MSG_RPBEFORE_CR: return "MSG_RPBEFORE_CR";
137 case H1_MSG_RPVER: return "MSG_RPVER";
138 case H1_MSG_RPVER_SP: return "MSG_RPVER_SP";
139 case H1_MSG_RPCODE: return "MSG_RPCODE";
140 case H1_MSG_RPCODE_SP: return "MSG_RPCODE_SP";
141 case H1_MSG_RPREASON: return "MSG_RPREASON";
142 case H1_MSG_RPLINE_END: return "MSG_RPLINE_END";
143 case H1_MSG_HDR_FIRST: return "MSG_HDR_FIRST";
144 case H1_MSG_HDR_NAME: return "MSG_HDR_NAME";
145 case H1_MSG_HDR_COL: return "MSG_HDR_COL";
146 case H1_MSG_HDR_L1_SP: return "MSG_HDR_L1_SP";
147 case H1_MSG_HDR_L1_LF: return "MSG_HDR_L1_LF";
148 case H1_MSG_HDR_L1_LWS: return "MSG_HDR_L1_LWS";
149 case H1_MSG_HDR_VAL: return "MSG_HDR_VAL";
150 case H1_MSG_HDR_L2_LF: return "MSG_HDR_L2_LF";
151 case H1_MSG_HDR_L2_LWS: return "MSG_HDR_L2_LWS";
152 case H1_MSG_LAST_LF: return "MSG_LAST_LF";
Willy Tarreau801250e2018-09-11 11:45:04 +0200153 case H1_MSG_CHUNK_SIZE: return "MSG_CHUNK_SIZE";
154 case H1_MSG_DATA: return "MSG_DATA";
155 case H1_MSG_CHUNK_CRLF: return "MSG_CHUNK_CRLF";
156 case H1_MSG_TRAILERS: return "MSG_TRAILERS";
Willy Tarreau801250e2018-09-11 11:45:04 +0200157 case H1_MSG_DONE: return "MSG_DONE";
Willy Tarreau801250e2018-09-11 11:45:04 +0200158 case H1_MSG_TUNNEL: return "MSG_TUNNEL";
159 default: return "MSG_??????";
160 }
161}
162
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200163/* This function may be called only in HTTP_MSG_CHUNK_CRLF. It reads the CRLF or
164 * a possible LF alone at the end of a chunk. The caller should adjust msg->next
165 * in order to include this part into the next forwarding phase. Note that the
Willy Tarreauc0973c62018-06-14 15:53:21 +0200166 * caller must ensure that head+start points to the first byte to parse. It
167 * returns the number of bytes parsed on success, so the caller can set msg_state
168 * to HTTP_MSG_CHUNK_SIZE. If not enough data are available, the function does not
Willy Tarreaub2892562017-09-21 11:33:54 +0200169 * change anything and returns zero. Otherwise it returns a negative value
170 * indicating the error positionn relative to <stop>. Note: this function is
171 * designed to parse wrapped CRLF at the end of the buffer.
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200172 */
Willy Tarreaub2892562017-09-21 11:33:54 +0200173static inline int h1_skip_chunk_crlf(const struct buffer *buf, int start, int stop)
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200174{
Willy Tarreauc0973c62018-06-14 15:53:21 +0200175 const char *ptr = b_peek(buf, start);
Willy Tarreaub2892562017-09-21 11:33:54 +0200176 int bytes = 1;
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200177
178 /* NB: we'll check data availabilty at the end. It's not a
179 * problem because whatever we match first will be checked
180 * against the correct length.
181 */
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200182 if (*ptr == '\r') {
183 bytes++;
184 ptr++;
Willy Tarreauc0973c62018-06-14 15:53:21 +0200185 if (ptr >= b_wrap(buf))
Willy Tarreau591d4452018-06-15 17:21:00 +0200186 ptr = b_orig(buf);
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200187 }
188
Willy Tarreaub2892562017-09-21 11:33:54 +0200189 if (bytes > stop - start)
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200190 return 0;
191
Willy Tarreauc0973c62018-06-14 15:53:21 +0200192 if (*ptr != '\n') // negative position to stop
193 return ptr - __b_peek(buf, stop);
Willy Tarreaub2892562017-09-21 11:33:54 +0200194
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200195 return bytes;
196}
197
Willy Tarreau84d6b7a2018-06-14 15:59:05 +0200198/* Parse the chunk size start at buf + start and stops before buf + stop. The
199 * positions are relative to the buffer's head.
Willy Tarreaue56cdd32017-09-21 08:36:33 +0200200 * It returns the chunk size in <res> and the amount of bytes read this way :
201 * < 0 : error at this position relative to <stop>
202 * = 0 : not enough bytes to read a complete chunk size
203 * > 0 : number of bytes successfully read that the caller can skip
204 * On success, the caller should adjust its msg->next to point to the first
205 * byte of data after the chunk size, so that we know we can forward exactly
206 * msg->next bytes, and msg->sol to contain the exact number of bytes forming
207 * the chunk size. That way it is always possible to differentiate between the
208 * start of the body and the start of the data. Note: this function is designed
209 * to parse wrapped CRLF at the end of the buffer.
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200210 */
Willy Tarreaue56cdd32017-09-21 08:36:33 +0200211static inline int h1_parse_chunk_size(const struct buffer *buf, int start, int stop, unsigned int *res)
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200212{
Willy Tarreau84d6b7a2018-06-14 15:59:05 +0200213 const char *ptr = b_peek(buf, start);
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200214 const char *ptr_old = ptr;
Willy Tarreau84d6b7a2018-06-14 15:59:05 +0200215 const char *end = b_wrap(buf);
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200216 unsigned int chunk = 0;
217
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100218 stop -= start; // bytes left
219 start = stop; // bytes to transfer
220
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200221 /* The chunk size is in the following form, though we are only
222 * interested in the size and CRLF :
223 * 1*HEXDIGIT *WSP *[ ';' extensions ] CRLF
224 */
225 while (1) {
226 int c;
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100227 if (!stop)
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200228 return 0;
229 c = hex2i(*ptr);
230 if (c < 0) /* not a hex digit anymore */
231 break;
232 if (unlikely(++ptr >= end))
Willy Tarreau591d4452018-06-15 17:21:00 +0200233 ptr = b_orig(buf);
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200234 if (unlikely(chunk & 0xF8000000)) /* integer overflow will occur if result >= 2GB */
235 goto error;
236 chunk = (chunk << 4) + c;
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100237 stop--;
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200238 }
239
240 /* empty size not allowed */
241 if (unlikely(ptr == ptr_old))
242 goto error;
243
244 while (HTTP_IS_SPHT(*ptr)) {
245 if (++ptr >= end)
Willy Tarreau591d4452018-06-15 17:21:00 +0200246 ptr = b_orig(buf);
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100247 if (--stop == 0)
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200248 return 0;
249 }
250
251 /* Up to there, we know that at least one byte is present at *ptr. Check
252 * for the end of chunk size.
253 */
254 while (1) {
255 if (likely(HTTP_IS_CRLF(*ptr))) {
256 /* we now have a CR or an LF at ptr */
257 if (likely(*ptr == '\r')) {
258 if (++ptr >= end)
Willy Tarreau591d4452018-06-15 17:21:00 +0200259 ptr = b_orig(buf);
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100260 if (--stop == 0)
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200261 return 0;
262 }
263
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100264 if (*ptr != '\n')
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200265 goto error;
266 if (++ptr >= end)
Willy Tarreau591d4452018-06-15 17:21:00 +0200267 ptr = b_orig(buf);
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100268 --stop;
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200269 /* done */
270 break;
271 }
272 else if (likely(*ptr == ';')) {
273 /* chunk extension, ends at next CRLF */
274 if (++ptr >= end)
Willy Tarreau591d4452018-06-15 17:21:00 +0200275 ptr = b_orig(buf);
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100276 if (--stop == 0)
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200277 return 0;
278
279 while (!HTTP_IS_CRLF(*ptr)) {
280 if (++ptr >= end)
Willy Tarreau591d4452018-06-15 17:21:00 +0200281 ptr = b_orig(buf);
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100282 if (--stop == 0)
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200283 return 0;
284 }
285 /* we have a CRLF now, loop above */
286 continue;
287 }
288 else
289 goto error;
290 }
291
292 /* OK we found our CRLF and now <ptr> points to the next byte, which may
Willy Tarreaue56cdd32017-09-21 08:36:33 +0200293 * or may not be present. Let's return the number of bytes parsed.
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200294 */
Willy Tarreaue56cdd32017-09-21 08:36:33 +0200295 *res = chunk;
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100296 return start - stop;
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200297 error:
Willy Tarreaue56cdd32017-09-21 08:36:33 +0200298 *res = 0; // just to stop gcc's -Wuninitialized warning :-(
Willy Tarreaub15e3fe2017-11-10 11:17:08 +0100299 return -stop;
Willy Tarreaudb4893d2017-09-21 08:40:02 +0200300}
301
Willy Tarreau7f437ff2018-09-11 13:51:19 +0200302/* initializes an H1 message for a request */
303static inline struct h1m *h1m_init_req(struct h1m *h1m)
Willy Tarreau4093a4d2017-09-21 11:46:43 +0200304{
Willy Tarreau801250e2018-09-11 11:45:04 +0200305 h1m->state = H1_MSG_RQBEFORE;
Willy Tarreaub3b01522018-09-11 11:51:31 +0200306 h1m->next = 0;
Willy Tarreauccaf2332018-09-11 16:47:23 +0200307 h1m->flags = H1_MF_NONE;
Willy Tarreau4093a4d2017-09-21 11:46:43 +0200308 h1m->curr_len = 0;
309 h1m->body_len = 0;
Willy Tarreaubbf38232018-09-12 09:08:54 +0200310 h1m->err_pos = -2;
Willy Tarreau4093a4d2017-09-21 11:46:43 +0200311 h1m->err_state = 0;
312 return h1m;
313}
Willy Tarreau0da5b3b2017-09-21 09:30:46 +0200314
Willy Tarreau7f437ff2018-09-11 13:51:19 +0200315/* initializes an H1 message for a response */
316static inline struct h1m *h1m_init_res(struct h1m *h1m)
317{
318 h1m->state = H1_MSG_RPBEFORE;
319 h1m->next = 0;
Willy Tarreauccaf2332018-09-11 16:47:23 +0200320 h1m->flags = H1_MF_RESP;
Willy Tarreau7f437ff2018-09-11 13:51:19 +0200321 h1m->curr_len = 0;
322 h1m->body_len = 0;
Willy Tarreaubbf38232018-09-12 09:08:54 +0200323 h1m->err_pos = -2;
Willy Tarreau7f437ff2018-09-11 13:51:19 +0200324 h1m->err_state = 0;
325 return h1m;
326}
327
Willy Tarreau0da5b3b2017-09-21 09:30:46 +0200328#endif /* _PROTO_H1_H */