/src/http-parser/http_parser.c
Line | Count | Source |
1 | | /* Copyright Joyent, Inc. and other Node contributors. |
2 | | * |
3 | | * Permission is hereby granted, free of charge, to any person obtaining a copy |
4 | | * of this software and associated documentation files (the "Software"), to |
5 | | * deal in the Software without restriction, including without limitation the |
6 | | * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
7 | | * sell copies of the Software, and to permit persons to whom the Software is |
8 | | * furnished to do so, subject to the following conditions: |
9 | | * |
10 | | * The above copyright notice and this permission notice shall be included in |
11 | | * all copies or substantial portions of the Software. |
12 | | * |
13 | | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
14 | | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
15 | | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
16 | | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
17 | | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
18 | | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS |
19 | | * IN THE SOFTWARE. |
20 | | */ |
21 | | #include "http_parser.h" |
22 | | #include <assert.h> |
23 | | #include <stddef.h> |
24 | | #include <ctype.h> |
25 | | #include <string.h> |
26 | | #include <limits.h> |
27 | | |
28 | | static uint32_t max_header_size = HTTP_MAX_HEADER_SIZE; |
29 | | |
30 | | #ifndef ULLONG_MAX |
31 | | # define ULLONG_MAX ((uint64_t) -1) /* 2^64-1 */ |
32 | | #endif |
33 | | |
34 | | #ifndef MIN |
35 | 0 | # define MIN(a,b) ((a) < (b) ? (a) : (b)) |
36 | | #endif |
37 | | |
38 | | #ifndef ARRAY_SIZE |
39 | 0 | # define ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0])) |
40 | | #endif |
41 | | |
42 | | #ifndef BIT_AT |
43 | | # define BIT_AT(a, i) \ |
44 | 1.68M | (!!((unsigned int) (a)[(unsigned int) (i) >> 3] & \ |
45 | 1.68M | (1 << ((unsigned int) (i) & 7)))) |
46 | | #endif |
47 | | |
48 | | #ifndef ELEM_AT |
49 | 0 | # define ELEM_AT(a, i, v) ((unsigned int) (i) < ARRAY_SIZE(a) ? (a)[(i)] : (v)) |
50 | | #endif |
51 | | |
52 | 0 | #define SET_ERRNO(e) \ |
53 | 0 | do { \ |
54 | 0 | parser->nread = nread; \ |
55 | 0 | parser->http_errno = (e); \ |
56 | 0 | } while(0) |
57 | | |
58 | 0 | #define CURRENT_STATE() p_state |
59 | 0 | #define UPDATE_STATE(V) p_state = (enum state) (V); |
60 | 0 | #define RETURN(V) \ |
61 | 0 | do { \ |
62 | 0 | parser->nread = nread; \ |
63 | 0 | parser->state = CURRENT_STATE(); \ |
64 | 0 | return (V); \ |
65 | 0 | } while (0); |
66 | | #define REEXECUTE() \ |
67 | 0 | goto reexecute; \ |
68 | | |
69 | | |
70 | | #ifdef __GNUC__ |
71 | 0 | # define LIKELY(X) __builtin_expect(!!(X), 1) |
72 | 0 | # define UNLIKELY(X) __builtin_expect(!!(X), 0) |
73 | | #else |
74 | | # define LIKELY(X) (X) |
75 | | # define UNLIKELY(X) (X) |
76 | | #endif |
77 | | |
78 | | |
79 | | /* Run the notify callback FOR, returning ER if it fails */ |
80 | 0 | #define CALLBACK_NOTIFY_(FOR, ER) \ |
81 | 0 | do { \ |
82 | 0 | assert(HTTP_PARSER_ERRNO(parser) == HPE_OK); \ |
83 | 0 | \ |
84 | 0 | if (LIKELY(settings->on_##FOR)) { \ |
85 | 0 | parser->state = CURRENT_STATE(); \ |
86 | 0 | if (UNLIKELY(0 != settings->on_##FOR(parser))) { \ |
87 | 0 | SET_ERRNO(HPE_CB_##FOR); \ |
88 | 0 | } \ |
89 | 0 | UPDATE_STATE(parser->state); \ |
90 | 0 | \ |
91 | 0 | /* We either errored above or got paused; get out */ \ |
92 | 0 | if (UNLIKELY(HTTP_PARSER_ERRNO(parser) != HPE_OK)) { \ |
93 | 0 | return (ER); \ |
94 | 0 | } \ |
95 | 0 | } \ |
96 | 0 | } while (0) |
97 | | |
98 | | /* Run the notify callback FOR and consume the current byte */ |
99 | 0 | #define CALLBACK_NOTIFY(FOR) CALLBACK_NOTIFY_(FOR, p - data + 1) |
100 | | |
101 | | /* Run the notify callback FOR and don't consume the current byte */ |
102 | 0 | #define CALLBACK_NOTIFY_NOADVANCE(FOR) CALLBACK_NOTIFY_(FOR, p - data) |
103 | | |
104 | | /* Run data callback FOR with LEN bytes, returning ER if it fails */ |
105 | 0 | #define CALLBACK_DATA_(FOR, LEN, ER) \ |
106 | 0 | do { \ |
107 | 0 | assert(HTTP_PARSER_ERRNO(parser) == HPE_OK); \ |
108 | 0 | \ |
109 | 0 | if (FOR##_mark) { \ |
110 | 0 | if (LIKELY(settings->on_##FOR)) { \ |
111 | 0 | parser->state = CURRENT_STATE(); \ |
112 | 0 | if (UNLIKELY(0 != \ |
113 | 0 | settings->on_##FOR(parser, FOR##_mark, (LEN)))) { \ |
114 | 0 | SET_ERRNO(HPE_CB_##FOR); \ |
115 | 0 | } \ |
116 | 0 | UPDATE_STATE(parser->state); \ |
117 | 0 | \ |
118 | 0 | /* We either errored above or got paused; get out */ \ |
119 | 0 | if (UNLIKELY(HTTP_PARSER_ERRNO(parser) != HPE_OK)) { \ |
120 | 0 | return (ER); \ |
121 | 0 | } \ |
122 | 0 | } \ |
123 | 0 | FOR##_mark = NULL; \ |
124 | 0 | } \ |
125 | 0 | } while (0) |
126 | | |
127 | | /* Run the data callback FOR and consume the current byte */ |
128 | | #define CALLBACK_DATA(FOR) \ |
129 | 0 | CALLBACK_DATA_(FOR, p - FOR##_mark, p - data + 1) |
130 | | |
131 | | /* Run the data callback FOR and don't consume the current byte */ |
132 | | #define CALLBACK_DATA_NOADVANCE(FOR) \ |
133 | 0 | CALLBACK_DATA_(FOR, p - FOR##_mark, p - data) |
134 | | |
135 | | /* Set the mark FOR; non-destructive if mark is already set */ |
136 | 0 | #define MARK(FOR) \ |
137 | 0 | do { \ |
138 | 0 | if (!FOR##_mark) { \ |
139 | 0 | FOR##_mark = p; \ |
140 | 0 | } \ |
141 | 0 | } while (0) |
142 | | |
143 | | /* Don't allow the total size of the HTTP headers (including the status |
144 | | * line) to exceed max_header_size. This check is here to protect |
145 | | * embedders against denial-of-service attacks where the attacker feeds |
146 | | * us a never-ending header that the embedder keeps buffering. |
147 | | * |
148 | | * This check is arguably the responsibility of embedders but we're doing |
149 | | * it on the embedder's behalf because most won't bother and this way we |
150 | | * make the web a little safer. max_header_size is still far bigger |
151 | | * than any reasonable request or response so this should never affect |
152 | | * day-to-day operation. |
153 | | */ |
154 | 0 | #define COUNT_HEADER_SIZE(V) \ |
155 | 0 | do { \ |
156 | 0 | nread += (uint32_t)(V); \ |
157 | 0 | if (UNLIKELY(nread > max_header_size)) { \ |
158 | 0 | SET_ERRNO(HPE_HEADER_OVERFLOW); \ |
159 | 0 | goto error; \ |
160 | 0 | } \ |
161 | 0 | } while (0) |
162 | | |
163 | | |
164 | 0 | #define PROXY_CONNECTION "proxy-connection" |
165 | 0 | #define CONNECTION "connection" |
166 | 0 | #define CONTENT_LENGTH "content-length" |
167 | 0 | #define TRANSFER_ENCODING "transfer-encoding" |
168 | 0 | #define UPGRADE "upgrade" |
169 | 0 | #define CHUNKED "chunked" |
170 | 0 | #define KEEP_ALIVE "keep-alive" |
171 | 0 | #define CLOSE "close" |
172 | | |
173 | | |
174 | | static const char *method_strings[] = |
175 | | { |
176 | | #define XX(num, name, string) #string, |
177 | | HTTP_METHOD_MAP(XX) |
178 | | #undef XX |
179 | | }; |
180 | | |
181 | | |
182 | | /* Tokens as defined by rfc 2616. Also lowercases them. |
183 | | * token = 1*<any CHAR except CTLs or separators> |
184 | | * separators = "(" | ")" | "<" | ">" | "@" |
185 | | * | "," | ";" | ":" | "\" | <"> |
186 | | * | "/" | "[" | "]" | "?" | "=" |
187 | | * | "{" | "}" | SP | HT |
188 | | */ |
189 | | static const char tokens[256] = { |
190 | | /* 0 nul 1 soh 2 stx 3 etx 4 eot 5 enq 6 ack 7 bel */ |
191 | | 0, 0, 0, 0, 0, 0, 0, 0, |
192 | | /* 8 bs 9 ht 10 nl 11 vt 12 np 13 cr 14 so 15 si */ |
193 | | 0, 0, 0, 0, 0, 0, 0, 0, |
194 | | /* 16 dle 17 dc1 18 dc2 19 dc3 20 dc4 21 nak 22 syn 23 etb */ |
195 | | 0, 0, 0, 0, 0, 0, 0, 0, |
196 | | /* 24 can 25 em 26 sub 27 esc 28 fs 29 gs 30 rs 31 us */ |
197 | | 0, 0, 0, 0, 0, 0, 0, 0, |
198 | | /* 32 sp 33 ! 34 " 35 # 36 $ 37 % 38 & 39 ' */ |
199 | | ' ', '!', 0, '#', '$', '%', '&', '\'', |
200 | | /* 40 ( 41 ) 42 * 43 + 44 , 45 - 46 . 47 / */ |
201 | | 0, 0, '*', '+', 0, '-', '.', 0, |
202 | | /* 48 0 49 1 50 2 51 3 52 4 53 5 54 6 55 7 */ |
203 | | '0', '1', '2', '3', '4', '5', '6', '7', |
204 | | /* 56 8 57 9 58 : 59 ; 60 < 61 = 62 > 63 ? */ |
205 | | '8', '9', 0, 0, 0, 0, 0, 0, |
206 | | /* 64 @ 65 A 66 B 67 C 68 D 69 E 70 F 71 G */ |
207 | | 0, 'a', 'b', 'c', 'd', 'e', 'f', 'g', |
208 | | /* 72 H 73 I 74 J 75 K 76 L 77 M 78 N 79 O */ |
209 | | 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', |
210 | | /* 80 P 81 Q 82 R 83 S 84 T 85 U 86 V 87 W */ |
211 | | 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', |
212 | | /* 88 X 89 Y 90 Z 91 [ 92 \ 93 ] 94 ^ 95 _ */ |
213 | | 'x', 'y', 'z', 0, 0, 0, '^', '_', |
214 | | /* 96 ` 97 a 98 b 99 c 100 d 101 e 102 f 103 g */ |
215 | | '`', 'a', 'b', 'c', 'd', 'e', 'f', 'g', |
216 | | /* 104 h 105 i 106 j 107 k 108 l 109 m 110 n 111 o */ |
217 | | 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', |
218 | | /* 112 p 113 q 114 r 115 s 116 t 117 u 118 v 119 w */ |
219 | | 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', |
220 | | /* 120 x 121 y 122 z 123 { 124 | 125 } 126 ~ 127 del */ |
221 | | 'x', 'y', 'z', 0, '|', 0, '~', 0 }; |
222 | | |
223 | | |
224 | | static const int8_t unhex[256] = |
225 | | {-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1 |
226 | | ,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1 |
227 | | ,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1 |
228 | | , 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,-1,-1,-1,-1,-1,-1 |
229 | | ,-1,10,11,12,13,14,15,-1,-1,-1,-1,-1,-1,-1,-1,-1 |
230 | | ,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1 |
231 | | ,-1,10,11,12,13,14,15,-1,-1,-1,-1,-1,-1,-1,-1,-1 |
232 | | ,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1 |
233 | | }; |
234 | | |
235 | | |
236 | | #if HTTP_PARSER_STRICT |
237 | | # define T(v) 0 |
238 | | #else |
239 | | # define T(v) v |
240 | | #endif |
241 | | |
242 | | |
243 | | static const uint8_t normal_url_char[32] = { |
244 | | /* 0 nul 1 soh 2 stx 3 etx 4 eot 5 enq 6 ack 7 bel */ |
245 | | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0, |
246 | | /* 8 bs 9 ht 10 nl 11 vt 12 np 13 cr 14 so 15 si */ |
247 | | 0 | T(2) | 0 | 0 | T(16) | 0 | 0 | 0, |
248 | | /* 16 dle 17 dc1 18 dc2 19 dc3 20 dc4 21 nak 22 syn 23 etb */ |
249 | | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0, |
250 | | /* 24 can 25 em 26 sub 27 esc 28 fs 29 gs 30 rs 31 us */ |
251 | | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0, |
252 | | /* 32 sp 33 ! 34 " 35 # 36 $ 37 % 38 & 39 ' */ |
253 | | 0 | 2 | 4 | 0 | 16 | 32 | 64 | 128, |
254 | | /* 40 ( 41 ) 42 * 43 + 44 , 45 - 46 . 47 / */ |
255 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
256 | | /* 48 0 49 1 50 2 51 3 52 4 53 5 54 6 55 7 */ |
257 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
258 | | /* 56 8 57 9 58 : 59 ; 60 < 61 = 62 > 63 ? */ |
259 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 0, |
260 | | /* 64 @ 65 A 66 B 67 C 68 D 69 E 70 F 71 G */ |
261 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
262 | | /* 72 H 73 I 74 J 75 K 76 L 77 M 78 N 79 O */ |
263 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
264 | | /* 80 P 81 Q 82 R 83 S 84 T 85 U 86 V 87 W */ |
265 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
266 | | /* 88 X 89 Y 90 Z 91 [ 92 \ 93 ] 94 ^ 95 _ */ |
267 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
268 | | /* 96 ` 97 a 98 b 99 c 100 d 101 e 102 f 103 g */ |
269 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
270 | | /* 104 h 105 i 106 j 107 k 108 l 109 m 110 n 111 o */ |
271 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
272 | | /* 112 p 113 q 114 r 115 s 116 t 117 u 118 v 119 w */ |
273 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128, |
274 | | /* 120 x 121 y 122 z 123 { 124 | 125 } 126 ~ 127 del */ |
275 | | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 0, }; |
276 | | |
277 | | #undef T |
278 | | |
279 | | enum state |
280 | | { s_dead = 1 /* important that this is > 0 */ |
281 | | |
282 | | , s_start_req_or_res |
283 | | , s_res_or_resp_H |
284 | | , s_start_res |
285 | | , s_res_H |
286 | | , s_res_HT |
287 | | , s_res_HTT |
288 | | , s_res_HTTP |
289 | | , s_res_http_major |
290 | | , s_res_http_dot |
291 | | , s_res_http_minor |
292 | | , s_res_http_end |
293 | | , s_res_first_status_code |
294 | | , s_res_status_code |
295 | | , s_res_status_start |
296 | | , s_res_status |
297 | | , s_res_line_almost_done |
298 | | |
299 | | , s_start_req |
300 | | |
301 | | , s_req_method |
302 | | , s_req_spaces_before_url |
303 | | , s_req_schema |
304 | | , s_req_schema_slash |
305 | | , s_req_schema_slash_slash |
306 | | , s_req_server_start |
307 | | , s_req_server |
308 | | , s_req_server_with_at |
309 | | , s_req_path |
310 | | , s_req_query_string_start |
311 | | , s_req_query_string |
312 | | , s_req_fragment_start |
313 | | , s_req_fragment |
314 | | , s_req_http_start |
315 | | , s_req_http_H |
316 | | , s_req_http_HT |
317 | | , s_req_http_HTT |
318 | | , s_req_http_HTTP |
319 | | , s_req_http_I |
320 | | , s_req_http_IC |
321 | | , s_req_http_major |
322 | | , s_req_http_dot |
323 | | , s_req_http_minor |
324 | | , s_req_http_end |
325 | | , s_req_line_almost_done |
326 | | |
327 | | , s_header_field_start |
328 | | , s_header_field |
329 | | , s_header_value_discard_ws |
330 | | , s_header_value_discard_ws_almost_done |
331 | | , s_header_value_discard_lws |
332 | | , s_header_value_start |
333 | | , s_header_value |
334 | | , s_header_value_lws |
335 | | |
336 | | , s_header_almost_done |
337 | | |
338 | | , s_chunk_size_start |
339 | | , s_chunk_size |
340 | | , s_chunk_parameters |
341 | | , s_chunk_size_almost_done |
342 | | |
343 | | , s_headers_almost_done |
344 | | , s_headers_done |
345 | | |
346 | | /* Important: 's_headers_done' must be the last 'header' state. All |
347 | | * states beyond this must be 'body' states. It is used for overflow |
348 | | * checking. See the PARSING_HEADER() macro. |
349 | | */ |
350 | | |
351 | | , s_chunk_data |
352 | | , s_chunk_data_almost_done |
353 | | , s_chunk_data_done |
354 | | |
355 | | , s_body_identity |
356 | | , s_body_identity_eof |
357 | | |
358 | | , s_message_done |
359 | | }; |
360 | | |
361 | | |
362 | 0 | #define PARSING_HEADER(state) (state <= s_headers_done) |
363 | | |
364 | | |
365 | | enum header_states |
366 | | { h_general = 0 |
367 | | , h_C |
368 | | , h_CO |
369 | | , h_CON |
370 | | |
371 | | , h_matching_connection |
372 | | , h_matching_proxy_connection |
373 | | , h_matching_content_length |
374 | | , h_matching_transfer_encoding |
375 | | , h_matching_upgrade |
376 | | |
377 | | , h_connection |
378 | | , h_content_length |
379 | | , h_content_length_num |
380 | | , h_content_length_ws |
381 | | , h_transfer_encoding |
382 | | , h_upgrade |
383 | | |
384 | | , h_matching_transfer_encoding_token_start |
385 | | , h_matching_transfer_encoding_chunked |
386 | | , h_matching_transfer_encoding_token |
387 | | |
388 | | , h_matching_connection_token_start |
389 | | , h_matching_connection_keep_alive |
390 | | , h_matching_connection_close |
391 | | , h_matching_connection_upgrade |
392 | | , h_matching_connection_token |
393 | | |
394 | | , h_transfer_encoding_chunked |
395 | | , h_connection_keep_alive |
396 | | , h_connection_close |
397 | | , h_connection_upgrade |
398 | | }; |
399 | | |
400 | | enum http_host_state |
401 | | { |
402 | | s_http_host_dead = 1 |
403 | | , s_http_userinfo_start |
404 | | , s_http_userinfo |
405 | | , s_http_host_start |
406 | | , s_http_host_v6_start |
407 | | , s_http_host |
408 | | , s_http_host_v6 |
409 | | , s_http_host_v6_end |
410 | | , s_http_host_v6_zone_start |
411 | | , s_http_host_v6_zone |
412 | | , s_http_host_port_start |
413 | | , s_http_host_port |
414 | | }; |
415 | | |
416 | | /* Macros for character classes; depends on strict-mode */ |
417 | 0 | #define CR '\r' |
418 | 0 | #define LF '\n' |
419 | 2.98M | #define LOWER(c) (unsigned char)(c | 0x20) |
420 | 3.11M | #define IS_ALPHA(c) (LOWER(c) >= 'a' && LOWER(c) <= 'z') |
421 | 486k | #define IS_NUM(c) ((c) >= '0' && (c) <= '9') |
422 | 2.82M | #define IS_ALPHANUM(c) (IS_ALPHA(c) || IS_NUM(c)) |
423 | 4.26k | #define IS_HEX(c) (IS_NUM(c) || (LOWER(c) >= 'a' && LOWER(c) <= 'f')) |
424 | 1.55M | #define IS_MARK(c) ((c) == '-' || (c) == '_' || (c) == '.' || \ |
425 | 362k | (c) == '!' || (c) == '~' || (c) == '*' || (c) == '\'' || (c) == '(' || \ |
426 | 1.55M | (c) == ')') |
427 | 2.09M | #define IS_USERINFO_CHAR(c) (IS_ALPHANUM(c) || IS_MARK(c) || (c) == '%' || \ |
428 | 1.19M | (c) == ';' || (c) == ':' || (c) == '&' || (c) == '=' || (c) == '+' || \ |
429 | 1.19M | (c) == '$' || (c) == ',') |
430 | | |
431 | 0 | #define STRICT_TOKEN(c) ((c == ' ') ? 0 : tokens[(unsigned char)c]) |
432 | | |
433 | | #if HTTP_PARSER_STRICT |
434 | | #define TOKEN(c) STRICT_TOKEN(c) |
435 | | #define IS_URL_CHAR(c) (BIT_AT(normal_url_char, (unsigned char)c)) |
436 | | #define IS_HOST_CHAR(c) (IS_ALPHANUM(c) || (c) == '.' || (c) == '-') |
437 | | #else |
438 | 0 | #define TOKEN(c) tokens[(unsigned char)c] |
439 | | #define IS_URL_CHAR(c) \ |
440 | 841k | (BIT_AT(normal_url_char, (unsigned char)c) || ((c) & 0x80)) |
441 | | #define IS_HOST_CHAR(c) \ |
442 | 221k | (IS_ALPHANUM(c) || (c) == '.' || (c) == '-' || (c) == '_') |
443 | | #endif |
444 | | |
445 | | /** |
446 | | * Verify that a char is a valid visible (printable) US-ASCII |
447 | | * character or %x80-FF |
448 | | **/ |
449 | | #define IS_HEADER_CHAR(ch) \ |
450 | 0 | (ch == CR || ch == LF || ch == 9 || ((unsigned char)ch > 31 && ch != 127)) |
451 | | |
452 | | #define start_state (parser->type == HTTP_REQUEST ? s_start_req : s_start_res) |
453 | | |
454 | | |
455 | | #if HTTP_PARSER_STRICT |
456 | | # define STRICT_CHECK(cond) \ |
457 | | do { \ |
458 | | if (cond) { \ |
459 | | SET_ERRNO(HPE_STRICT); \ |
460 | | goto error; \ |
461 | | } \ |
462 | | } while (0) |
463 | | # define NEW_MESSAGE() (http_should_keep_alive(parser) ? start_state : s_dead) |
464 | | #else |
465 | | # define STRICT_CHECK(cond) |
466 | | # define NEW_MESSAGE() start_state |
467 | | #endif |
468 | | |
469 | | |
470 | | /* Map errno values to strings for human-readable output */ |
471 | | #define HTTP_STRERROR_GEN(n, s) { "HPE_" #n, s }, |
472 | | static struct { |
473 | | const char *name; |
474 | | const char *description; |
475 | | } http_strerror_tab[] = { |
476 | | HTTP_ERRNO_MAP(HTTP_STRERROR_GEN) |
477 | | }; |
478 | | #undef HTTP_STRERROR_GEN |
479 | | |
480 | | int http_message_needs_eof(const http_parser *parser); |
481 | | |
482 | | /* Our URL parser. |
483 | | * |
484 | | * This is designed to be shared by http_parser_execute() for URL validation, |
485 | | * hence it has a state transition + byte-for-byte interface. In addition, it |
486 | | * is meant to be embedded in http_parser_parse_url(), which does the dirty |
487 | | * work of turning state transitions URL components for its API. |
488 | | * |
489 | | * This function should only be invoked with non-space characters. It is |
490 | | * assumed that the caller cares about (and can detect) the transition between |
491 | | * URL and non-URL states by looking for these. |
492 | | */ |
493 | | static enum state |
494 | | parse_url_char(enum state s, const char ch) |
495 | 2.03M | { |
496 | 2.03M | if (ch == ' ' || ch == '\r' || ch == '\n') { |
497 | 9 | return s_dead; |
498 | 9 | } |
499 | | |
500 | | #if HTTP_PARSER_STRICT |
501 | | if (ch == '\t' || ch == '\f') { |
502 | | return s_dead; |
503 | | } |
504 | | #endif |
505 | | |
506 | 2.03M | switch (s) { |
507 | 963 | case s_req_spaces_before_url: |
508 | | /* Proxied requests are followed by scheme of an absolute URI (alpha). |
509 | | * All methods except CONNECT are followed by '/' or '*'. |
510 | | */ |
511 | | |
512 | 963 | if (ch == '/' || ch == '*') { |
513 | 143 | return s_req_path; |
514 | 143 | } |
515 | | |
516 | 820 | if (IS_ALPHA(ch)) { |
517 | 165 | return s_req_schema; |
518 | 165 | } |
519 | | |
520 | 655 | break; |
521 | | |
522 | 284k | case s_req_schema: |
523 | 284k | if (IS_ALPHA(ch)) { |
524 | 284k | return s; |
525 | 284k | } |
526 | | |
527 | 144 | if (ch == ':') { |
528 | 81 | return s_req_schema_slash; |
529 | 81 | } |
530 | | |
531 | 63 | break; |
532 | | |
533 | 80 | case s_req_schema_slash: |
534 | 80 | if (ch == '/') { |
535 | 37 | return s_req_schema_slash_slash; |
536 | 37 | } |
537 | | |
538 | 43 | break; |
539 | | |
540 | 43 | case s_req_schema_slash_slash: |
541 | 36 | if (ch == '/') { |
542 | 27 | return s_req_server_start; |
543 | 27 | } |
544 | | |
545 | 9 | break; |
546 | | |
547 | 266 | case s_req_server_with_at: |
548 | 266 | if (ch == '@') { |
549 | 1 | return s_dead; |
550 | 1 | } |
551 | | |
552 | | /* fall through */ |
553 | 1.25k | case s_req_server_start: |
554 | 906k | case s_req_server: |
555 | 906k | if (ch == '/') { |
556 | 139 | return s_req_path; |
557 | 139 | } |
558 | | |
559 | 906k | if (ch == '?') { |
560 | 78 | return s_req_query_string_start; |
561 | 78 | } |
562 | | |
563 | 906k | if (ch == '@') { |
564 | 450 | return s_req_server_with_at; |
565 | 450 | } |
566 | | |
567 | 905k | if (IS_USERINFO_CHAR(ch) || ch == '[' || ch == ']') { |
568 | 905k | return s_req_server; |
569 | 905k | } |
570 | | |
571 | 88 | break; |
572 | | |
573 | 404k | case s_req_path: |
574 | 404k | if (IS_URL_CHAR(ch)) { |
575 | 404k | return s; |
576 | 404k | } |
577 | | |
578 | 186 | switch (ch) { |
579 | 55 | case '?': |
580 | 55 | return s_req_query_string_start; |
581 | | |
582 | 113 | case '#': |
583 | 113 | return s_req_fragment_start; |
584 | 186 | } |
585 | | |
586 | 18 | break; |
587 | | |
588 | 98 | case s_req_query_string_start: |
589 | 436k | case s_req_query_string: |
590 | 436k | if (IS_URL_CHAR(ch)) { |
591 | 435k | return s_req_query_string; |
592 | 435k | } |
593 | | |
594 | 289 | switch (ch) { |
595 | 238 | case '?': |
596 | | /* allow extra '?' in query string */ |
597 | 238 | return s_req_query_string; |
598 | | |
599 | 25 | case '#': |
600 | 25 | return s_req_fragment_start; |
601 | 289 | } |
602 | | |
603 | 26 | break; |
604 | | |
605 | 483 | case s_req_fragment_start: |
606 | 483 | if (IS_URL_CHAR(ch)) { |
607 | 88 | return s_req_fragment; |
608 | 88 | } |
609 | | |
610 | 395 | switch (ch) { |
611 | 9 | case '?': |
612 | 9 | return s_req_fragment; |
613 | | |
614 | 364 | case '#': |
615 | 364 | return s; |
616 | 395 | } |
617 | | |
618 | 22 | break; |
619 | | |
620 | 783 | case s_req_fragment: |
621 | 783 | if (IS_URL_CHAR(ch)) { |
622 | 574 | return s; |
623 | 574 | } |
624 | | |
625 | 209 | switch (ch) { |
626 | 48 | case '?': |
627 | 191 | case '#': |
628 | 191 | return s; |
629 | 209 | } |
630 | | |
631 | 18 | break; |
632 | | |
633 | 18 | default: |
634 | 0 | break; |
635 | 2.03M | } |
636 | | |
637 | | /* We should never fall out of the switch above unless there's an error */ |
638 | 942 | return s_dead; |
639 | 2.03M | } |
640 | | |
641 | | size_t http_parser_execute (http_parser *parser, |
642 | | const http_parser_settings *settings, |
643 | | const char *data, |
644 | | size_t len) |
645 | 0 | { |
646 | 0 | char c, ch; |
647 | 0 | int8_t unhex_val; |
648 | 0 | const char *p = data; |
649 | 0 | const char *header_field_mark = 0; |
650 | 0 | const char *header_value_mark = 0; |
651 | 0 | const char *url_mark = 0; |
652 | 0 | const char *body_mark = 0; |
653 | 0 | const char *status_mark = 0; |
654 | 0 | enum state p_state = (enum state) parser->state; |
655 | 0 | const unsigned int lenient = parser->lenient_http_headers; |
656 | 0 | const unsigned int allow_chunked_length = parser->allow_chunked_length; |
657 | |
|
658 | 0 | uint32_t nread = parser->nread; |
659 | | |
660 | | /* We're in an error state. Don't bother doing anything. */ |
661 | 0 | if (HTTP_PARSER_ERRNO(parser) != HPE_OK) { |
662 | 0 | return 0; |
663 | 0 | } |
664 | | |
665 | 0 | if (len == 0) { |
666 | 0 | switch (CURRENT_STATE()) { |
667 | 0 | case s_body_identity_eof: |
668 | | /* Use of CALLBACK_NOTIFY() here would erroneously return 1 byte read if |
669 | | * we got paused. |
670 | | */ |
671 | 0 | CALLBACK_NOTIFY_NOADVANCE(message_complete); |
672 | 0 | return 0; |
673 | | |
674 | 0 | case s_dead: |
675 | 0 | case s_start_req_or_res: |
676 | 0 | case s_start_res: |
677 | 0 | case s_start_req: |
678 | 0 | return 0; |
679 | | |
680 | 0 | default: |
681 | 0 | SET_ERRNO(HPE_INVALID_EOF_STATE); |
682 | 0 | return 1; |
683 | 0 | } |
684 | 0 | } |
685 | | |
686 | | |
687 | 0 | if (CURRENT_STATE() == s_header_field) |
688 | 0 | header_field_mark = data; |
689 | 0 | if (CURRENT_STATE() == s_header_value) |
690 | 0 | header_value_mark = data; |
691 | 0 | switch (CURRENT_STATE()) { |
692 | 0 | case s_req_path: |
693 | 0 | case s_req_schema: |
694 | 0 | case s_req_schema_slash: |
695 | 0 | case s_req_schema_slash_slash: |
696 | 0 | case s_req_server_start: |
697 | 0 | case s_req_server: |
698 | 0 | case s_req_server_with_at: |
699 | 0 | case s_req_query_string_start: |
700 | 0 | case s_req_query_string: |
701 | 0 | case s_req_fragment_start: |
702 | 0 | case s_req_fragment: |
703 | 0 | url_mark = data; |
704 | 0 | break; |
705 | 0 | case s_res_status: |
706 | 0 | status_mark = data; |
707 | 0 | break; |
708 | 0 | default: |
709 | 0 | break; |
710 | 0 | } |
711 | | |
712 | 0 | for (p=data; p != data + len; p++) { |
713 | 0 | ch = *p; |
714 | |
|
715 | 0 | if (PARSING_HEADER(CURRENT_STATE())) |
716 | 0 | COUNT_HEADER_SIZE(1); |
717 | | |
718 | 0 | reexecute: |
719 | 0 | switch (CURRENT_STATE()) { |
720 | | |
721 | 0 | case s_dead: |
722 | | /* this state is used after a 'Connection: close' message |
723 | | * the parser will error out if it reads another message |
724 | | */ |
725 | 0 | if (LIKELY(ch == CR || ch == LF)) |
726 | 0 | break; |
727 | | |
728 | 0 | SET_ERRNO(HPE_CLOSED_CONNECTION); |
729 | 0 | goto error; |
730 | | |
731 | 0 | case s_start_req_or_res: |
732 | 0 | { |
733 | 0 | if (ch == CR || ch == LF) |
734 | 0 | break; |
735 | 0 | parser->flags = 0; |
736 | 0 | parser->uses_transfer_encoding = 0; |
737 | 0 | parser->content_length = ULLONG_MAX; |
738 | |
|
739 | 0 | if (ch == 'H') { |
740 | 0 | UPDATE_STATE(s_res_or_resp_H); |
741 | |
|
742 | 0 | CALLBACK_NOTIFY(message_begin); |
743 | 0 | } else { |
744 | 0 | parser->type = HTTP_REQUEST; |
745 | 0 | UPDATE_STATE(s_start_req); |
746 | 0 | REEXECUTE(); |
747 | 0 | } |
748 | | |
749 | 0 | break; |
750 | 0 | } |
751 | | |
752 | 0 | case s_res_or_resp_H: |
753 | 0 | if (ch == 'T') { |
754 | 0 | parser->type = HTTP_RESPONSE; |
755 | 0 | UPDATE_STATE(s_res_HT); |
756 | 0 | } else { |
757 | 0 | if (UNLIKELY(ch != 'E')) { |
758 | 0 | SET_ERRNO(HPE_INVALID_CONSTANT); |
759 | 0 | goto error; |
760 | 0 | } |
761 | | |
762 | 0 | parser->type = HTTP_REQUEST; |
763 | 0 | parser->method = HTTP_HEAD; |
764 | 0 | parser->index = 2; |
765 | 0 | UPDATE_STATE(s_req_method); |
766 | 0 | } |
767 | 0 | break; |
768 | | |
769 | 0 | case s_start_res: |
770 | 0 | { |
771 | 0 | if (ch == CR || ch == LF) |
772 | 0 | break; |
773 | 0 | parser->flags = 0; |
774 | 0 | parser->uses_transfer_encoding = 0; |
775 | 0 | parser->content_length = ULLONG_MAX; |
776 | |
|
777 | 0 | if (ch == 'H') { |
778 | 0 | UPDATE_STATE(s_res_H); |
779 | 0 | } else { |
780 | 0 | SET_ERRNO(HPE_INVALID_CONSTANT); |
781 | 0 | goto error; |
782 | 0 | } |
783 | | |
784 | 0 | CALLBACK_NOTIFY(message_begin); |
785 | 0 | break; |
786 | 0 | } |
787 | | |
788 | 0 | case s_res_H: |
789 | 0 | STRICT_CHECK(ch != 'T'); |
790 | 0 | UPDATE_STATE(s_res_HT); |
791 | 0 | break; |
792 | | |
793 | 0 | case s_res_HT: |
794 | 0 | STRICT_CHECK(ch != 'T'); |
795 | 0 | UPDATE_STATE(s_res_HTT); |
796 | 0 | break; |
797 | | |
798 | 0 | case s_res_HTT: |
799 | 0 | STRICT_CHECK(ch != 'P'); |
800 | 0 | UPDATE_STATE(s_res_HTTP); |
801 | 0 | break; |
802 | | |
803 | 0 | case s_res_HTTP: |
804 | 0 | STRICT_CHECK(ch != '/'); |
805 | 0 | UPDATE_STATE(s_res_http_major); |
806 | 0 | break; |
807 | | |
808 | 0 | case s_res_http_major: |
809 | 0 | if (UNLIKELY(!IS_NUM(ch))) { |
810 | 0 | SET_ERRNO(HPE_INVALID_VERSION); |
811 | 0 | goto error; |
812 | 0 | } |
813 | | |
814 | 0 | parser->http_major = ch - '0'; |
815 | 0 | UPDATE_STATE(s_res_http_dot); |
816 | 0 | break; |
817 | | |
818 | 0 | case s_res_http_dot: |
819 | 0 | { |
820 | 0 | if (UNLIKELY(ch != '.')) { |
821 | 0 | SET_ERRNO(HPE_INVALID_VERSION); |
822 | 0 | goto error; |
823 | 0 | } |
824 | | |
825 | 0 | UPDATE_STATE(s_res_http_minor); |
826 | 0 | break; |
827 | 0 | } |
828 | | |
829 | 0 | case s_res_http_minor: |
830 | 0 | if (UNLIKELY(!IS_NUM(ch))) { |
831 | 0 | SET_ERRNO(HPE_INVALID_VERSION); |
832 | 0 | goto error; |
833 | 0 | } |
834 | | |
835 | 0 | parser->http_minor = ch - '0'; |
836 | 0 | UPDATE_STATE(s_res_http_end); |
837 | 0 | break; |
838 | | |
839 | 0 | case s_res_http_end: |
840 | 0 | { |
841 | 0 | if (UNLIKELY(ch != ' ')) { |
842 | 0 | SET_ERRNO(HPE_INVALID_VERSION); |
843 | 0 | goto error; |
844 | 0 | } |
845 | | |
846 | 0 | UPDATE_STATE(s_res_first_status_code); |
847 | 0 | break; |
848 | 0 | } |
849 | | |
850 | 0 | case s_res_first_status_code: |
851 | 0 | { |
852 | 0 | if (!IS_NUM(ch)) { |
853 | 0 | if (ch == ' ') { |
854 | 0 | break; |
855 | 0 | } |
856 | | |
857 | 0 | SET_ERRNO(HPE_INVALID_STATUS); |
858 | 0 | goto error; |
859 | 0 | } |
860 | 0 | parser->status_code = ch - '0'; |
861 | 0 | UPDATE_STATE(s_res_status_code); |
862 | 0 | break; |
863 | 0 | } |
864 | | |
865 | 0 | case s_res_status_code: |
866 | 0 | { |
867 | 0 | if (!IS_NUM(ch)) { |
868 | 0 | switch (ch) { |
869 | 0 | case ' ': |
870 | 0 | UPDATE_STATE(s_res_status_start); |
871 | 0 | break; |
872 | 0 | case CR: |
873 | 0 | case LF: |
874 | 0 | UPDATE_STATE(s_res_status_start); |
875 | 0 | REEXECUTE(); |
876 | 0 | break; |
877 | 0 | default: |
878 | 0 | SET_ERRNO(HPE_INVALID_STATUS); |
879 | 0 | goto error; |
880 | 0 | } |
881 | 0 | break; |
882 | 0 | } |
883 | | |
884 | 0 | parser->status_code *= 10; |
885 | 0 | parser->status_code += ch - '0'; |
886 | |
|
887 | 0 | if (UNLIKELY(parser->status_code > 999)) { |
888 | 0 | SET_ERRNO(HPE_INVALID_STATUS); |
889 | 0 | goto error; |
890 | 0 | } |
891 | | |
892 | 0 | break; |
893 | 0 | } |
894 | | |
895 | 0 | case s_res_status_start: |
896 | 0 | { |
897 | 0 | MARK(status); |
898 | 0 | UPDATE_STATE(s_res_status); |
899 | 0 | parser->index = 0; |
900 | |
|
901 | 0 | if (ch == CR || ch == LF) |
902 | 0 | REEXECUTE(); |
903 | |
|
904 | 0 | break; |
905 | 0 | } |
906 | | |
907 | 0 | case s_res_status: |
908 | 0 | if (ch == CR) { |
909 | 0 | UPDATE_STATE(s_res_line_almost_done); |
910 | 0 | CALLBACK_DATA(status); |
911 | 0 | break; |
912 | 0 | } |
913 | | |
914 | 0 | if (ch == LF) { |
915 | 0 | UPDATE_STATE(s_header_field_start); |
916 | 0 | CALLBACK_DATA(status); |
917 | 0 | break; |
918 | 0 | } |
919 | | |
920 | 0 | break; |
921 | | |
922 | 0 | case s_res_line_almost_done: |
923 | 0 | STRICT_CHECK(ch != LF); |
924 | 0 | UPDATE_STATE(s_header_field_start); |
925 | 0 | break; |
926 | | |
927 | 0 | case s_start_req: |
928 | 0 | { |
929 | 0 | if (ch == CR || ch == LF) |
930 | 0 | break; |
931 | 0 | parser->flags = 0; |
932 | 0 | parser->uses_transfer_encoding = 0; |
933 | 0 | parser->content_length = ULLONG_MAX; |
934 | |
|
935 | 0 | if (UNLIKELY(!IS_ALPHA(ch))) { |
936 | 0 | SET_ERRNO(HPE_INVALID_METHOD); |
937 | 0 | goto error; |
938 | 0 | } |
939 | | |
940 | 0 | parser->method = (enum http_method) 0; |
941 | 0 | parser->index = 1; |
942 | 0 | switch (ch) { |
943 | 0 | case 'A': parser->method = HTTP_ACL; break; |
944 | 0 | case 'B': parser->method = HTTP_BIND; break; |
945 | 0 | case 'C': parser->method = HTTP_CONNECT; /* or COPY, CHECKOUT */ break; |
946 | 0 | case 'D': parser->method = HTTP_DELETE; break; |
947 | 0 | case 'G': parser->method = HTTP_GET; break; |
948 | 0 | case 'H': parser->method = HTTP_HEAD; break; |
949 | 0 | case 'L': parser->method = HTTP_LOCK; /* or LINK */ break; |
950 | 0 | case 'M': parser->method = HTTP_MKCOL; /* or MOVE, MKACTIVITY, MERGE, M-SEARCH, MKCALENDAR */ break; |
951 | 0 | case 'N': parser->method = HTTP_NOTIFY; break; |
952 | 0 | case 'O': parser->method = HTTP_OPTIONS; break; |
953 | 0 | case 'P': parser->method = HTTP_POST; |
954 | | /* or PROPFIND|PROPPATCH|PUT|PATCH|PURGE */ |
955 | 0 | break; |
956 | 0 | case 'R': parser->method = HTTP_REPORT; /* or REBIND */ break; |
957 | 0 | case 'S': parser->method = HTTP_SUBSCRIBE; /* or SEARCH, SOURCE */ break; |
958 | 0 | case 'T': parser->method = HTTP_TRACE; break; |
959 | 0 | case 'U': parser->method = HTTP_UNLOCK; /* or UNSUBSCRIBE, UNBIND, UNLINK */ break; |
960 | 0 | default: |
961 | 0 | SET_ERRNO(HPE_INVALID_METHOD); |
962 | 0 | goto error; |
963 | 0 | } |
964 | 0 | UPDATE_STATE(s_req_method); |
965 | |
|
966 | 0 | CALLBACK_NOTIFY(message_begin); |
967 | | |
968 | 0 | break; |
969 | 0 | } |
970 | | |
971 | 0 | case s_req_method: |
972 | 0 | { |
973 | 0 | const char *matcher; |
974 | 0 | if (UNLIKELY(ch == '\0')) { |
975 | 0 | SET_ERRNO(HPE_INVALID_METHOD); |
976 | 0 | goto error; |
977 | 0 | } |
978 | | |
979 | 0 | matcher = method_strings[parser->method]; |
980 | 0 | if (ch == ' ' && matcher[parser->index] == '\0') { |
981 | 0 | UPDATE_STATE(s_req_spaces_before_url); |
982 | 0 | } else if (ch == matcher[parser->index]) { |
983 | 0 | ; /* nada */ |
984 | 0 | } else if ((ch >= 'A' && ch <= 'Z') || ch == '-') { |
985 | |
|
986 | 0 | switch (parser->method << 16 | parser->index << 8 | ch) { |
987 | 0 | #define XX(meth, pos, ch, new_meth) \ |
988 | 0 | case (HTTP_##meth << 16 | pos << 8 | ch): \ |
989 | 0 | parser->method = HTTP_##new_meth; break; |
990 | | |
991 | 0 | XX(POST, 1, 'U', PUT) |
992 | 0 | XX(POST, 1, 'A', PATCH) |
993 | 0 | XX(POST, 1, 'R', PROPFIND) |
994 | 0 | XX(PUT, 2, 'R', PURGE) |
995 | 0 | XX(CONNECT, 1, 'H', CHECKOUT) |
996 | 0 | XX(CONNECT, 2, 'P', COPY) |
997 | 0 | XX(MKCOL, 1, 'O', MOVE) |
998 | 0 | XX(MKCOL, 1, 'E', MERGE) |
999 | 0 | XX(MKCOL, 1, '-', MSEARCH) |
1000 | 0 | XX(MKCOL, 2, 'A', MKACTIVITY) |
1001 | 0 | XX(MKCOL, 3, 'A', MKCALENDAR) |
1002 | 0 | XX(SUBSCRIBE, 1, 'E', SEARCH) |
1003 | 0 | XX(SUBSCRIBE, 1, 'O', SOURCE) |
1004 | 0 | XX(REPORT, 2, 'B', REBIND) |
1005 | 0 | XX(PROPFIND, 4, 'P', PROPPATCH) |
1006 | 0 | XX(LOCK, 1, 'I', LINK) |
1007 | 0 | XX(UNLOCK, 2, 'S', UNSUBSCRIBE) |
1008 | 0 | XX(UNLOCK, 2, 'B', UNBIND) |
1009 | 0 | XX(UNLOCK, 3, 'I', UNLINK) |
1010 | 0 | #undef XX |
1011 | 0 | default: |
1012 | 0 | SET_ERRNO(HPE_INVALID_METHOD); |
1013 | 0 | goto error; |
1014 | 0 | } |
1015 | 0 | } else { |
1016 | 0 | SET_ERRNO(HPE_INVALID_METHOD); |
1017 | 0 | goto error; |
1018 | 0 | } |
1019 | | |
1020 | 0 | ++parser->index; |
1021 | 0 | break; |
1022 | 0 | } |
1023 | | |
1024 | 0 | case s_req_spaces_before_url: |
1025 | 0 | { |
1026 | 0 | if (ch == ' ') break; |
1027 | | |
1028 | 0 | MARK(url); |
1029 | 0 | if (parser->method == HTTP_CONNECT) { |
1030 | 0 | UPDATE_STATE(s_req_server_start); |
1031 | 0 | } |
1032 | |
|
1033 | 0 | UPDATE_STATE(parse_url_char(CURRENT_STATE(), ch)); |
1034 | 0 | if (UNLIKELY(CURRENT_STATE() == s_dead)) { |
1035 | 0 | SET_ERRNO(HPE_INVALID_URL); |
1036 | 0 | goto error; |
1037 | 0 | } |
1038 | | |
1039 | 0 | break; |
1040 | 0 | } |
1041 | | |
1042 | 0 | case s_req_schema: |
1043 | 0 | case s_req_schema_slash: |
1044 | 0 | case s_req_schema_slash_slash: |
1045 | 0 | case s_req_server_start: |
1046 | 0 | { |
1047 | 0 | switch (ch) { |
1048 | | /* No whitespace allowed here */ |
1049 | 0 | case ' ': |
1050 | 0 | case CR: |
1051 | 0 | case LF: |
1052 | 0 | SET_ERRNO(HPE_INVALID_URL); |
1053 | 0 | goto error; |
1054 | 0 | default: |
1055 | 0 | UPDATE_STATE(parse_url_char(CURRENT_STATE(), ch)); |
1056 | 0 | if (UNLIKELY(CURRENT_STATE() == s_dead)) { |
1057 | 0 | SET_ERRNO(HPE_INVALID_URL); |
1058 | 0 | goto error; |
1059 | 0 | } |
1060 | 0 | } |
1061 | | |
1062 | 0 | break; |
1063 | 0 | } |
1064 | | |
1065 | 0 | case s_req_server: |
1066 | 0 | case s_req_server_with_at: |
1067 | 0 | case s_req_path: |
1068 | 0 | case s_req_query_string_start: |
1069 | 0 | case s_req_query_string: |
1070 | 0 | case s_req_fragment_start: |
1071 | 0 | case s_req_fragment: |
1072 | 0 | { |
1073 | 0 | switch (ch) { |
1074 | 0 | case ' ': |
1075 | 0 | UPDATE_STATE(s_req_http_start); |
1076 | 0 | CALLBACK_DATA(url); |
1077 | 0 | break; |
1078 | 0 | case CR: |
1079 | 0 | case LF: |
1080 | 0 | parser->http_major = 0; |
1081 | 0 | parser->http_minor = 9; |
1082 | 0 | UPDATE_STATE((ch == CR) ? |
1083 | 0 | s_req_line_almost_done : |
1084 | 0 | s_header_field_start); |
1085 | 0 | CALLBACK_DATA(url); |
1086 | 0 | break; |
1087 | 0 | default: |
1088 | 0 | UPDATE_STATE(parse_url_char(CURRENT_STATE(), ch)); |
1089 | 0 | if (UNLIKELY(CURRENT_STATE() == s_dead)) { |
1090 | 0 | SET_ERRNO(HPE_INVALID_URL); |
1091 | 0 | goto error; |
1092 | 0 | } |
1093 | 0 | } |
1094 | 0 | break; |
1095 | 0 | } |
1096 | | |
1097 | 0 | case s_req_http_start: |
1098 | 0 | switch (ch) { |
1099 | 0 | case ' ': |
1100 | 0 | break; |
1101 | 0 | case 'H': |
1102 | 0 | UPDATE_STATE(s_req_http_H); |
1103 | 0 | break; |
1104 | 0 | case 'I': |
1105 | 0 | if (parser->method == HTTP_SOURCE) { |
1106 | 0 | UPDATE_STATE(s_req_http_I); |
1107 | 0 | break; |
1108 | 0 | } |
1109 | | /* fall through */ |
1110 | 0 | default: |
1111 | 0 | SET_ERRNO(HPE_INVALID_CONSTANT); |
1112 | 0 | goto error; |
1113 | 0 | } |
1114 | 0 | break; |
1115 | | |
1116 | 0 | case s_req_http_H: |
1117 | 0 | STRICT_CHECK(ch != 'T'); |
1118 | 0 | UPDATE_STATE(s_req_http_HT); |
1119 | 0 | break; |
1120 | | |
1121 | 0 | case s_req_http_HT: |
1122 | 0 | STRICT_CHECK(ch != 'T'); |
1123 | 0 | UPDATE_STATE(s_req_http_HTT); |
1124 | 0 | break; |
1125 | | |
1126 | 0 | case s_req_http_HTT: |
1127 | 0 | STRICT_CHECK(ch != 'P'); |
1128 | 0 | UPDATE_STATE(s_req_http_HTTP); |
1129 | 0 | break; |
1130 | | |
1131 | 0 | case s_req_http_I: |
1132 | 0 | STRICT_CHECK(ch != 'C'); |
1133 | 0 | UPDATE_STATE(s_req_http_IC); |
1134 | 0 | break; |
1135 | | |
1136 | 0 | case s_req_http_IC: |
1137 | 0 | STRICT_CHECK(ch != 'E'); |
1138 | 0 | UPDATE_STATE(s_req_http_HTTP); /* Treat "ICE" as "HTTP". */ |
1139 | 0 | break; |
1140 | | |
1141 | 0 | case s_req_http_HTTP: |
1142 | 0 | STRICT_CHECK(ch != '/'); |
1143 | 0 | UPDATE_STATE(s_req_http_major); |
1144 | 0 | break; |
1145 | | |
1146 | 0 | case s_req_http_major: |
1147 | 0 | if (UNLIKELY(!IS_NUM(ch))) { |
1148 | 0 | SET_ERRNO(HPE_INVALID_VERSION); |
1149 | 0 | goto error; |
1150 | 0 | } |
1151 | | |
1152 | 0 | parser->http_major = ch - '0'; |
1153 | 0 | UPDATE_STATE(s_req_http_dot); |
1154 | 0 | break; |
1155 | | |
1156 | 0 | case s_req_http_dot: |
1157 | 0 | { |
1158 | 0 | if (UNLIKELY(ch != '.')) { |
1159 | 0 | SET_ERRNO(HPE_INVALID_VERSION); |
1160 | 0 | goto error; |
1161 | 0 | } |
1162 | | |
1163 | 0 | UPDATE_STATE(s_req_http_minor); |
1164 | 0 | break; |
1165 | 0 | } |
1166 | | |
1167 | 0 | case s_req_http_minor: |
1168 | 0 | if (UNLIKELY(!IS_NUM(ch))) { |
1169 | 0 | SET_ERRNO(HPE_INVALID_VERSION); |
1170 | 0 | goto error; |
1171 | 0 | } |
1172 | | |
1173 | 0 | parser->http_minor = ch - '0'; |
1174 | 0 | UPDATE_STATE(s_req_http_end); |
1175 | 0 | break; |
1176 | | |
1177 | 0 | case s_req_http_end: |
1178 | 0 | { |
1179 | 0 | if (ch == CR) { |
1180 | 0 | UPDATE_STATE(s_req_line_almost_done); |
1181 | 0 | break; |
1182 | 0 | } |
1183 | | |
1184 | 0 | if (ch == LF) { |
1185 | 0 | UPDATE_STATE(s_header_field_start); |
1186 | 0 | break; |
1187 | 0 | } |
1188 | | |
1189 | 0 | SET_ERRNO(HPE_INVALID_VERSION); |
1190 | 0 | goto error; |
1191 | 0 | break; |
1192 | 0 | } |
1193 | | |
1194 | | /* end of request line */ |
1195 | 0 | case s_req_line_almost_done: |
1196 | 0 | { |
1197 | 0 | if (UNLIKELY(ch != LF)) { |
1198 | 0 | SET_ERRNO(HPE_LF_EXPECTED); |
1199 | 0 | goto error; |
1200 | 0 | } |
1201 | | |
1202 | 0 | UPDATE_STATE(s_header_field_start); |
1203 | 0 | break; |
1204 | 0 | } |
1205 | | |
1206 | 0 | case s_header_field_start: |
1207 | 0 | { |
1208 | 0 | if (ch == CR) { |
1209 | 0 | UPDATE_STATE(s_headers_almost_done); |
1210 | 0 | break; |
1211 | 0 | } |
1212 | | |
1213 | 0 | if (ch == LF) { |
1214 | | /* they might be just sending \n instead of \r\n so this would be |
1215 | | * the second \n to denote the end of headers*/ |
1216 | 0 | UPDATE_STATE(s_headers_almost_done); |
1217 | 0 | REEXECUTE(); |
1218 | 0 | } |
1219 | | |
1220 | 0 | c = TOKEN(ch); |
1221 | |
|
1222 | 0 | if (UNLIKELY(!c)) { |
1223 | 0 | SET_ERRNO(HPE_INVALID_HEADER_TOKEN); |
1224 | 0 | goto error; |
1225 | 0 | } |
1226 | | |
1227 | 0 | MARK(header_field); |
1228 | |
|
1229 | 0 | parser->index = 0; |
1230 | 0 | UPDATE_STATE(s_header_field); |
1231 | |
|
1232 | 0 | switch (c) { |
1233 | 0 | case 'c': |
1234 | 0 | parser->header_state = h_C; |
1235 | 0 | break; |
1236 | | |
1237 | 0 | case 'p': |
1238 | 0 | parser->header_state = h_matching_proxy_connection; |
1239 | 0 | break; |
1240 | | |
1241 | 0 | case 't': |
1242 | 0 | parser->header_state = h_matching_transfer_encoding; |
1243 | 0 | break; |
1244 | | |
1245 | 0 | case 'u': |
1246 | 0 | parser->header_state = h_matching_upgrade; |
1247 | 0 | break; |
1248 | | |
1249 | 0 | default: |
1250 | 0 | parser->header_state = h_general; |
1251 | 0 | break; |
1252 | 0 | } |
1253 | 0 | break; |
1254 | 0 | } |
1255 | | |
1256 | 0 | case s_header_field: |
1257 | 0 | { |
1258 | 0 | const char* start = p; |
1259 | 0 | for (; p != data + len; p++) { |
1260 | 0 | ch = *p; |
1261 | 0 | c = TOKEN(ch); |
1262 | |
|
1263 | 0 | if (!c) |
1264 | 0 | break; |
1265 | | |
1266 | 0 | switch (parser->header_state) { |
1267 | 0 | case h_general: { |
1268 | 0 | size_t left = data + len - p; |
1269 | 0 | const char* pe = p + MIN(left, max_header_size); |
1270 | 0 | while (p+1 < pe && TOKEN(p[1])) { |
1271 | 0 | p++; |
1272 | 0 | } |
1273 | 0 | break; |
1274 | 0 | } |
1275 | | |
1276 | 0 | case h_C: |
1277 | 0 | parser->index++; |
1278 | 0 | parser->header_state = (c == 'o' ? h_CO : h_general); |
1279 | 0 | break; |
1280 | | |
1281 | 0 | case h_CO: |
1282 | 0 | parser->index++; |
1283 | 0 | parser->header_state = (c == 'n' ? h_CON : h_general); |
1284 | 0 | break; |
1285 | | |
1286 | 0 | case h_CON: |
1287 | 0 | parser->index++; |
1288 | 0 | switch (c) { |
1289 | 0 | case 'n': |
1290 | 0 | parser->header_state = h_matching_connection; |
1291 | 0 | break; |
1292 | 0 | case 't': |
1293 | 0 | parser->header_state = h_matching_content_length; |
1294 | 0 | break; |
1295 | 0 | default: |
1296 | 0 | parser->header_state = h_general; |
1297 | 0 | break; |
1298 | 0 | } |
1299 | 0 | break; |
1300 | | |
1301 | | /* connection */ |
1302 | | |
1303 | 0 | case h_matching_connection: |
1304 | 0 | parser->index++; |
1305 | 0 | if (parser->index > sizeof(CONNECTION)-1 |
1306 | 0 | || c != CONNECTION[parser->index]) { |
1307 | 0 | parser->header_state = h_general; |
1308 | 0 | } else if (parser->index == sizeof(CONNECTION)-2) { |
1309 | 0 | parser->header_state = h_connection; |
1310 | 0 | } |
1311 | 0 | break; |
1312 | | |
1313 | | /* proxy-connection */ |
1314 | | |
1315 | 0 | case h_matching_proxy_connection: |
1316 | 0 | parser->index++; |
1317 | 0 | if (parser->index > sizeof(PROXY_CONNECTION)-1 |
1318 | 0 | || c != PROXY_CONNECTION[parser->index]) { |
1319 | 0 | parser->header_state = h_general; |
1320 | 0 | } else if (parser->index == sizeof(PROXY_CONNECTION)-2) { |
1321 | 0 | parser->header_state = h_connection; |
1322 | 0 | } |
1323 | 0 | break; |
1324 | | |
1325 | | /* content-length */ |
1326 | | |
1327 | 0 | case h_matching_content_length: |
1328 | 0 | parser->index++; |
1329 | 0 | if (parser->index > sizeof(CONTENT_LENGTH)-1 |
1330 | 0 | || c != CONTENT_LENGTH[parser->index]) { |
1331 | 0 | parser->header_state = h_general; |
1332 | 0 | } else if (parser->index == sizeof(CONTENT_LENGTH)-2) { |
1333 | 0 | parser->header_state = h_content_length; |
1334 | 0 | } |
1335 | 0 | break; |
1336 | | |
1337 | | /* transfer-encoding */ |
1338 | | |
1339 | 0 | case h_matching_transfer_encoding: |
1340 | 0 | parser->index++; |
1341 | 0 | if (parser->index > sizeof(TRANSFER_ENCODING)-1 |
1342 | 0 | || c != TRANSFER_ENCODING[parser->index]) { |
1343 | 0 | parser->header_state = h_general; |
1344 | 0 | } else if (parser->index == sizeof(TRANSFER_ENCODING)-2) { |
1345 | 0 | parser->header_state = h_transfer_encoding; |
1346 | 0 | parser->uses_transfer_encoding = 1; |
1347 | 0 | } |
1348 | 0 | break; |
1349 | | |
1350 | | /* upgrade */ |
1351 | | |
1352 | 0 | case h_matching_upgrade: |
1353 | 0 | parser->index++; |
1354 | 0 | if (parser->index > sizeof(UPGRADE)-1 |
1355 | 0 | || c != UPGRADE[parser->index]) { |
1356 | 0 | parser->header_state = h_general; |
1357 | 0 | } else if (parser->index == sizeof(UPGRADE)-2) { |
1358 | 0 | parser->header_state = h_upgrade; |
1359 | 0 | } |
1360 | 0 | break; |
1361 | | |
1362 | 0 | case h_connection: |
1363 | 0 | case h_content_length: |
1364 | 0 | case h_transfer_encoding: |
1365 | 0 | case h_upgrade: |
1366 | 0 | if (ch != ' ') parser->header_state = h_general; |
1367 | 0 | break; |
1368 | | |
1369 | 0 | default: |
1370 | 0 | assert(0 && "Unknown header_state"); |
1371 | 0 | break; |
1372 | 0 | } |
1373 | 0 | } |
1374 | | |
1375 | 0 | if (p == data + len) { |
1376 | 0 | --p; |
1377 | 0 | COUNT_HEADER_SIZE(p - start); |
1378 | 0 | break; |
1379 | 0 | } |
1380 | | |
1381 | 0 | COUNT_HEADER_SIZE(p - start); |
1382 | | |
1383 | 0 | if (ch == ':') { |
1384 | 0 | UPDATE_STATE(s_header_value_discard_ws); |
1385 | 0 | CALLBACK_DATA(header_field); |
1386 | 0 | break; |
1387 | 0 | } |
1388 | | |
1389 | 0 | SET_ERRNO(HPE_INVALID_HEADER_TOKEN); |
1390 | 0 | goto error; |
1391 | 0 | } |
1392 | | |
1393 | 0 | case s_header_value_discard_ws: |
1394 | 0 | if (ch == ' ' || ch == '\t') break; |
1395 | | |
1396 | 0 | if (ch == CR) { |
1397 | 0 | UPDATE_STATE(s_header_value_discard_ws_almost_done); |
1398 | 0 | break; |
1399 | 0 | } |
1400 | | |
1401 | 0 | if (ch == LF) { |
1402 | 0 | UPDATE_STATE(s_header_value_discard_lws); |
1403 | 0 | break; |
1404 | 0 | } |
1405 | | |
1406 | | /* fall through */ |
1407 | | |
1408 | 0 | case s_header_value_start: |
1409 | 0 | { |
1410 | 0 | MARK(header_value); |
1411 | |
|
1412 | 0 | UPDATE_STATE(s_header_value); |
1413 | 0 | parser->index = 0; |
1414 | |
|
1415 | 0 | c = LOWER(ch); |
1416 | |
|
1417 | 0 | switch (parser->header_state) { |
1418 | 0 | case h_upgrade: |
1419 | 0 | parser->flags |= F_UPGRADE; |
1420 | 0 | parser->header_state = h_general; |
1421 | 0 | break; |
1422 | | |
1423 | 0 | case h_transfer_encoding: |
1424 | | /* looking for 'Transfer-Encoding: chunked' */ |
1425 | 0 | if ('c' == c) { |
1426 | 0 | parser->header_state = h_matching_transfer_encoding_chunked; |
1427 | 0 | } else { |
1428 | 0 | parser->header_state = h_matching_transfer_encoding_token; |
1429 | 0 | } |
1430 | 0 | break; |
1431 | | |
1432 | | /* Multi-value `Transfer-Encoding` header */ |
1433 | 0 | case h_matching_transfer_encoding_token_start: |
1434 | 0 | break; |
1435 | | |
1436 | 0 | case h_content_length: |
1437 | 0 | if (UNLIKELY(!IS_NUM(ch))) { |
1438 | 0 | SET_ERRNO(HPE_INVALID_CONTENT_LENGTH); |
1439 | 0 | goto error; |
1440 | 0 | } |
1441 | | |
1442 | 0 | if (parser->flags & F_CONTENTLENGTH) { |
1443 | 0 | SET_ERRNO(HPE_UNEXPECTED_CONTENT_LENGTH); |
1444 | 0 | goto error; |
1445 | 0 | } |
1446 | | |
1447 | 0 | parser->flags |= F_CONTENTLENGTH; |
1448 | 0 | parser->content_length = ch - '0'; |
1449 | 0 | parser->header_state = h_content_length_num; |
1450 | 0 | break; |
1451 | | |
1452 | | /* when obsolete line folding is encountered for content length |
1453 | | * continue to the s_header_value state */ |
1454 | 0 | case h_content_length_ws: |
1455 | 0 | break; |
1456 | | |
1457 | 0 | case h_connection: |
1458 | | /* looking for 'Connection: keep-alive' */ |
1459 | 0 | if (c == 'k') { |
1460 | 0 | parser->header_state = h_matching_connection_keep_alive; |
1461 | | /* looking for 'Connection: close' */ |
1462 | 0 | } else if (c == 'c') { |
1463 | 0 | parser->header_state = h_matching_connection_close; |
1464 | 0 | } else if (c == 'u') { |
1465 | 0 | parser->header_state = h_matching_connection_upgrade; |
1466 | 0 | } else { |
1467 | 0 | parser->header_state = h_matching_connection_token; |
1468 | 0 | } |
1469 | 0 | break; |
1470 | | |
1471 | | /* Multi-value `Connection` header */ |
1472 | 0 | case h_matching_connection_token_start: |
1473 | 0 | break; |
1474 | | |
1475 | 0 | default: |
1476 | 0 | parser->header_state = h_general; |
1477 | 0 | break; |
1478 | 0 | } |
1479 | 0 | break; |
1480 | 0 | } |
1481 | | |
1482 | 0 | case s_header_value: |
1483 | 0 | { |
1484 | 0 | const char* start = p; |
1485 | 0 | enum header_states h_state = (enum header_states) parser->header_state; |
1486 | 0 | for (; p != data + len; p++) { |
1487 | 0 | ch = *p; |
1488 | 0 | if (ch == CR) { |
1489 | 0 | UPDATE_STATE(s_header_almost_done); |
1490 | 0 | parser->header_state = h_state; |
1491 | 0 | CALLBACK_DATA(header_value); |
1492 | 0 | break; |
1493 | 0 | } |
1494 | | |
1495 | 0 | if (ch == LF) { |
1496 | 0 | UPDATE_STATE(s_header_almost_done); |
1497 | 0 | COUNT_HEADER_SIZE(p - start); |
1498 | 0 | parser->header_state = h_state; |
1499 | 0 | CALLBACK_DATA_NOADVANCE(header_value); |
1500 | 0 | REEXECUTE(); |
1501 | 0 | } |
1502 | | |
1503 | 0 | if (!lenient && !IS_HEADER_CHAR(ch)) { |
1504 | 0 | SET_ERRNO(HPE_INVALID_HEADER_TOKEN); |
1505 | 0 | goto error; |
1506 | 0 | } |
1507 | | |
1508 | 0 | c = LOWER(ch); |
1509 | |
|
1510 | 0 | switch (h_state) { |
1511 | 0 | case h_general: |
1512 | 0 | { |
1513 | 0 | size_t left = data + len - p; |
1514 | 0 | const char* pe = p + MIN(left, max_header_size); |
1515 | |
|
1516 | 0 | for (; p != pe; p++) { |
1517 | 0 | ch = *p; |
1518 | 0 | if (ch == CR || ch == LF) { |
1519 | 0 | --p; |
1520 | 0 | break; |
1521 | 0 | } |
1522 | 0 | if (!lenient && !IS_HEADER_CHAR(ch)) { |
1523 | 0 | SET_ERRNO(HPE_INVALID_HEADER_TOKEN); |
1524 | 0 | goto error; |
1525 | 0 | } |
1526 | 0 | } |
1527 | 0 | if (p == data + len) |
1528 | 0 | --p; |
1529 | 0 | break; |
1530 | 0 | } |
1531 | | |
1532 | 0 | case h_connection: |
1533 | 0 | case h_transfer_encoding: |
1534 | 0 | assert(0 && "Shouldn't get here."); |
1535 | 0 | break; |
1536 | | |
1537 | 0 | case h_content_length: |
1538 | 0 | if (ch == ' ') break; |
1539 | 0 | h_state = h_content_length_num; |
1540 | | /* fall through */ |
1541 | |
|
1542 | 0 | case h_content_length_num: |
1543 | 0 | { |
1544 | 0 | uint64_t t; |
1545 | |
|
1546 | 0 | if (ch == ' ') { |
1547 | 0 | h_state = h_content_length_ws; |
1548 | 0 | break; |
1549 | 0 | } |
1550 | | |
1551 | 0 | if (UNLIKELY(!IS_NUM(ch))) { |
1552 | 0 | SET_ERRNO(HPE_INVALID_CONTENT_LENGTH); |
1553 | 0 | parser->header_state = h_state; |
1554 | 0 | goto error; |
1555 | 0 | } |
1556 | | |
1557 | 0 | t = parser->content_length; |
1558 | 0 | t *= 10; |
1559 | 0 | t += ch - '0'; |
1560 | | |
1561 | | /* Overflow? Test against a conservative limit for simplicity. */ |
1562 | 0 | if (UNLIKELY((ULLONG_MAX - 10) / 10 < parser->content_length)) { |
1563 | 0 | SET_ERRNO(HPE_INVALID_CONTENT_LENGTH); |
1564 | 0 | parser->header_state = h_state; |
1565 | 0 | goto error; |
1566 | 0 | } |
1567 | | |
1568 | 0 | parser->content_length = t; |
1569 | 0 | break; |
1570 | 0 | } |
1571 | | |
1572 | 0 | case h_content_length_ws: |
1573 | 0 | if (ch == ' ') break; |
1574 | 0 | SET_ERRNO(HPE_INVALID_CONTENT_LENGTH); |
1575 | 0 | parser->header_state = h_state; |
1576 | 0 | goto error; |
1577 | | |
1578 | | /* Transfer-Encoding: chunked */ |
1579 | 0 | case h_matching_transfer_encoding_token_start: |
1580 | | /* looking for 'Transfer-Encoding: chunked' */ |
1581 | 0 | if ('c' == c) { |
1582 | 0 | h_state = h_matching_transfer_encoding_chunked; |
1583 | 0 | } else if (STRICT_TOKEN(c)) { |
1584 | | /* TODO(indutny): similar code below does this, but why? |
1585 | | * At the very least it seems to be inconsistent given that |
1586 | | * h_matching_transfer_encoding_token does not check for |
1587 | | * `STRICT_TOKEN` |
1588 | | */ |
1589 | 0 | h_state = h_matching_transfer_encoding_token; |
1590 | 0 | } else if (c == ' ' || c == '\t') { |
1591 | | /* Skip lws */ |
1592 | 0 | } else { |
1593 | 0 | h_state = h_general; |
1594 | 0 | } |
1595 | 0 | break; |
1596 | | |
1597 | 0 | case h_matching_transfer_encoding_chunked: |
1598 | 0 | parser->index++; |
1599 | 0 | if (parser->index > sizeof(CHUNKED)-1 |
1600 | 0 | || c != CHUNKED[parser->index]) { |
1601 | 0 | h_state = h_matching_transfer_encoding_token; |
1602 | 0 | } else if (parser->index == sizeof(CHUNKED)-2) { |
1603 | 0 | h_state = h_transfer_encoding_chunked; |
1604 | 0 | } |
1605 | 0 | break; |
1606 | | |
1607 | 0 | case h_matching_transfer_encoding_token: |
1608 | 0 | if (ch == ',') { |
1609 | 0 | h_state = h_matching_transfer_encoding_token_start; |
1610 | 0 | parser->index = 0; |
1611 | 0 | } |
1612 | 0 | break; |
1613 | | |
1614 | 0 | case h_matching_connection_token_start: |
1615 | | /* looking for 'Connection: keep-alive' */ |
1616 | 0 | if (c == 'k') { |
1617 | 0 | h_state = h_matching_connection_keep_alive; |
1618 | | /* looking for 'Connection: close' */ |
1619 | 0 | } else if (c == 'c') { |
1620 | 0 | h_state = h_matching_connection_close; |
1621 | 0 | } else if (c == 'u') { |
1622 | 0 | h_state = h_matching_connection_upgrade; |
1623 | 0 | } else if (STRICT_TOKEN(c)) { |
1624 | 0 | h_state = h_matching_connection_token; |
1625 | 0 | } else if (c == ' ' || c == '\t') { |
1626 | | /* Skip lws */ |
1627 | 0 | } else { |
1628 | 0 | h_state = h_general; |
1629 | 0 | } |
1630 | 0 | break; |
1631 | | |
1632 | | /* looking for 'Connection: keep-alive' */ |
1633 | 0 | case h_matching_connection_keep_alive: |
1634 | 0 | parser->index++; |
1635 | 0 | if (parser->index > sizeof(KEEP_ALIVE)-1 |
1636 | 0 | || c != KEEP_ALIVE[parser->index]) { |
1637 | 0 | h_state = h_matching_connection_token; |
1638 | 0 | } else if (parser->index == sizeof(KEEP_ALIVE)-2) { |
1639 | 0 | h_state = h_connection_keep_alive; |
1640 | 0 | } |
1641 | 0 | break; |
1642 | | |
1643 | | /* looking for 'Connection: close' */ |
1644 | 0 | case h_matching_connection_close: |
1645 | 0 | parser->index++; |
1646 | 0 | if (parser->index > sizeof(CLOSE)-1 || c != CLOSE[parser->index]) { |
1647 | 0 | h_state = h_matching_connection_token; |
1648 | 0 | } else if (parser->index == sizeof(CLOSE)-2) { |
1649 | 0 | h_state = h_connection_close; |
1650 | 0 | } |
1651 | 0 | break; |
1652 | | |
1653 | | /* looking for 'Connection: upgrade' */ |
1654 | 0 | case h_matching_connection_upgrade: |
1655 | 0 | parser->index++; |
1656 | 0 | if (parser->index > sizeof(UPGRADE) - 1 || |
1657 | 0 | c != UPGRADE[parser->index]) { |
1658 | 0 | h_state = h_matching_connection_token; |
1659 | 0 | } else if (parser->index == sizeof(UPGRADE)-2) { |
1660 | 0 | h_state = h_connection_upgrade; |
1661 | 0 | } |
1662 | 0 | break; |
1663 | | |
1664 | 0 | case h_matching_connection_token: |
1665 | 0 | if (ch == ',') { |
1666 | 0 | h_state = h_matching_connection_token_start; |
1667 | 0 | parser->index = 0; |
1668 | 0 | } |
1669 | 0 | break; |
1670 | | |
1671 | 0 | case h_transfer_encoding_chunked: |
1672 | 0 | if (ch != ' ') h_state = h_matching_transfer_encoding_token; |
1673 | 0 | break; |
1674 | | |
1675 | 0 | case h_connection_keep_alive: |
1676 | 0 | case h_connection_close: |
1677 | 0 | case h_connection_upgrade: |
1678 | 0 | if (ch == ',') { |
1679 | 0 | if (h_state == h_connection_keep_alive) { |
1680 | 0 | parser->flags |= F_CONNECTION_KEEP_ALIVE; |
1681 | 0 | } else if (h_state == h_connection_close) { |
1682 | 0 | parser->flags |= F_CONNECTION_CLOSE; |
1683 | 0 | } else if (h_state == h_connection_upgrade) { |
1684 | 0 | parser->flags |= F_CONNECTION_UPGRADE; |
1685 | 0 | } |
1686 | 0 | h_state = h_matching_connection_token_start; |
1687 | 0 | parser->index = 0; |
1688 | 0 | } else if (ch != ' ') { |
1689 | 0 | h_state = h_matching_connection_token; |
1690 | 0 | } |
1691 | 0 | break; |
1692 | | |
1693 | 0 | default: |
1694 | 0 | UPDATE_STATE(s_header_value); |
1695 | 0 | h_state = h_general; |
1696 | 0 | break; |
1697 | 0 | } |
1698 | 0 | } |
1699 | 0 | parser->header_state = h_state; |
1700 | |
|
1701 | 0 | if (p == data + len) |
1702 | 0 | --p; |
1703 | |
|
1704 | 0 | COUNT_HEADER_SIZE(p - start); |
1705 | 0 | break; |
1706 | 0 | } |
1707 | | |
1708 | 0 | case s_header_almost_done: |
1709 | 0 | { |
1710 | 0 | if (UNLIKELY(ch != LF)) { |
1711 | 0 | SET_ERRNO(HPE_LF_EXPECTED); |
1712 | 0 | goto error; |
1713 | 0 | } |
1714 | | |
1715 | 0 | UPDATE_STATE(s_header_value_lws); |
1716 | 0 | break; |
1717 | 0 | } |
1718 | | |
1719 | 0 | case s_header_value_lws: |
1720 | 0 | { |
1721 | 0 | if (ch == ' ' || ch == '\t') { |
1722 | 0 | if (parser->header_state == h_content_length_num) { |
1723 | | /* treat obsolete line folding as space */ |
1724 | 0 | parser->header_state = h_content_length_ws; |
1725 | 0 | } |
1726 | 0 | UPDATE_STATE(s_header_value_start); |
1727 | 0 | REEXECUTE(); |
1728 | 0 | } |
1729 | | |
1730 | | /* finished the header */ |
1731 | 0 | switch (parser->header_state) { |
1732 | 0 | case h_connection_keep_alive: |
1733 | 0 | parser->flags |= F_CONNECTION_KEEP_ALIVE; |
1734 | 0 | break; |
1735 | 0 | case h_connection_close: |
1736 | 0 | parser->flags |= F_CONNECTION_CLOSE; |
1737 | 0 | break; |
1738 | 0 | case h_transfer_encoding_chunked: |
1739 | 0 | parser->flags |= F_CHUNKED; |
1740 | 0 | break; |
1741 | 0 | case h_connection_upgrade: |
1742 | 0 | parser->flags |= F_CONNECTION_UPGRADE; |
1743 | 0 | break; |
1744 | 0 | default: |
1745 | 0 | break; |
1746 | 0 | } |
1747 | | |
1748 | 0 | UPDATE_STATE(s_header_field_start); |
1749 | 0 | REEXECUTE(); |
1750 | 0 | } |
1751 | | |
1752 | 0 | case s_header_value_discard_ws_almost_done: |
1753 | 0 | { |
1754 | 0 | STRICT_CHECK(ch != LF); |
1755 | 0 | UPDATE_STATE(s_header_value_discard_lws); |
1756 | 0 | break; |
1757 | 0 | } |
1758 | | |
1759 | 0 | case s_header_value_discard_lws: |
1760 | 0 | { |
1761 | 0 | if (ch == ' ' || ch == '\t') { |
1762 | 0 | UPDATE_STATE(s_header_value_discard_ws); |
1763 | 0 | break; |
1764 | 0 | } else { |
1765 | 0 | switch (parser->header_state) { |
1766 | 0 | case h_connection_keep_alive: |
1767 | 0 | parser->flags |= F_CONNECTION_KEEP_ALIVE; |
1768 | 0 | break; |
1769 | 0 | case h_connection_close: |
1770 | 0 | parser->flags |= F_CONNECTION_CLOSE; |
1771 | 0 | break; |
1772 | 0 | case h_connection_upgrade: |
1773 | 0 | parser->flags |= F_CONNECTION_UPGRADE; |
1774 | 0 | break; |
1775 | 0 | case h_transfer_encoding_chunked: |
1776 | 0 | parser->flags |= F_CHUNKED; |
1777 | 0 | break; |
1778 | 0 | case h_content_length: |
1779 | | /* do not allow empty content length */ |
1780 | 0 | SET_ERRNO(HPE_INVALID_CONTENT_LENGTH); |
1781 | 0 | goto error; |
1782 | 0 | break; |
1783 | 0 | default: |
1784 | 0 | break; |
1785 | 0 | } |
1786 | | |
1787 | | /* header value was empty */ |
1788 | 0 | MARK(header_value); |
1789 | 0 | UPDATE_STATE(s_header_field_start); |
1790 | 0 | CALLBACK_DATA_NOADVANCE(header_value); |
1791 | 0 | REEXECUTE(); |
1792 | 0 | } |
1793 | 0 | } |
1794 | | |
1795 | 0 | case s_headers_almost_done: |
1796 | 0 | { |
1797 | 0 | STRICT_CHECK(ch != LF); |
1798 | |
|
1799 | 0 | if (parser->flags & F_TRAILING) { |
1800 | | /* End of a chunked request */ |
1801 | 0 | UPDATE_STATE(s_message_done); |
1802 | 0 | CALLBACK_NOTIFY_NOADVANCE(chunk_complete); |
1803 | 0 | REEXECUTE(); |
1804 | 0 | } |
1805 | | |
1806 | | /* Cannot use transfer-encoding and a content-length header together |
1807 | | per the HTTP specification. (RFC 7230 Section 3.3.3) */ |
1808 | 0 | if ((parser->uses_transfer_encoding == 1) && |
1809 | 0 | (parser->flags & F_CONTENTLENGTH)) { |
1810 | | /* Allow it for lenient parsing as long as `Transfer-Encoding` is |
1811 | | * not `chunked` or allow_length_with_encoding is set |
1812 | | */ |
1813 | 0 | if (parser->flags & F_CHUNKED) { |
1814 | 0 | if (!allow_chunked_length) { |
1815 | 0 | SET_ERRNO(HPE_UNEXPECTED_CONTENT_LENGTH); |
1816 | 0 | goto error; |
1817 | 0 | } |
1818 | 0 | } else if (!lenient) { |
1819 | 0 | SET_ERRNO(HPE_UNEXPECTED_CONTENT_LENGTH); |
1820 | 0 | goto error; |
1821 | 0 | } |
1822 | 0 | } |
1823 | | |
1824 | 0 | UPDATE_STATE(s_headers_done); |
1825 | | |
1826 | | /* Set this here so that on_headers_complete() callbacks can see it */ |
1827 | 0 | if ((parser->flags & F_UPGRADE) && |
1828 | 0 | (parser->flags & F_CONNECTION_UPGRADE)) { |
1829 | | /* For responses, "Upgrade: foo" and "Connection: upgrade" are |
1830 | | * mandatory only when it is a 101 Switching Protocols response, |
1831 | | * otherwise it is purely informational, to announce support. |
1832 | | */ |
1833 | 0 | parser->upgrade = |
1834 | 0 | (parser->type == HTTP_REQUEST || parser->status_code == 101); |
1835 | 0 | } else { |
1836 | 0 | parser->upgrade = (parser->method == HTTP_CONNECT); |
1837 | 0 | } |
1838 | | |
1839 | | /* Here we call the headers_complete callback. This is somewhat |
1840 | | * different than other callbacks because if the user returns 1, we |
1841 | | * will interpret that as saying that this message has no body. This |
1842 | | * is needed for the annoying case of recieving a response to a HEAD |
1843 | | * request. |
1844 | | * |
1845 | | * We'd like to use CALLBACK_NOTIFY_NOADVANCE() here but we cannot, so |
1846 | | * we have to simulate it by handling a change in errno below. |
1847 | | */ |
1848 | 0 | if (settings->on_headers_complete) { |
1849 | 0 | switch (settings->on_headers_complete(parser)) { |
1850 | 0 | case 0: |
1851 | 0 | break; |
1852 | | |
1853 | 0 | case 2: |
1854 | 0 | parser->upgrade = 1; |
1855 | | |
1856 | | /* fall through */ |
1857 | 0 | case 1: |
1858 | 0 | parser->flags |= F_SKIPBODY; |
1859 | 0 | break; |
1860 | | |
1861 | 0 | default: |
1862 | 0 | SET_ERRNO(HPE_CB_headers_complete); |
1863 | 0 | RETURN(p - data); /* Error */ |
1864 | 0 | } |
1865 | 0 | } |
1866 | | |
1867 | 0 | if (HTTP_PARSER_ERRNO(parser) != HPE_OK) { |
1868 | 0 | RETURN(p - data); |
1869 | 0 | } |
1870 | | |
1871 | 0 | REEXECUTE(); |
1872 | 0 | } |
1873 | | |
1874 | 0 | case s_headers_done: |
1875 | 0 | { |
1876 | 0 | int hasBody; |
1877 | 0 | STRICT_CHECK(ch != LF); |
1878 | |
|
1879 | 0 | parser->nread = 0; |
1880 | 0 | nread = 0; |
1881 | |
|
1882 | 0 | hasBody = parser->flags & F_CHUNKED || |
1883 | 0 | (parser->content_length > 0 && parser->content_length != ULLONG_MAX); |
1884 | 0 | if (parser->upgrade && (parser->method == HTTP_CONNECT || |
1885 | 0 | (parser->flags & F_SKIPBODY) || !hasBody)) { |
1886 | | /* Exit, the rest of the message is in a different protocol. */ |
1887 | 0 | UPDATE_STATE(NEW_MESSAGE()); |
1888 | 0 | CALLBACK_NOTIFY(message_complete); |
1889 | 0 | RETURN((p - data) + 1); |
1890 | 0 | } |
1891 | | |
1892 | 0 | if (parser->flags & F_SKIPBODY) { |
1893 | 0 | UPDATE_STATE(NEW_MESSAGE()); |
1894 | 0 | CALLBACK_NOTIFY(message_complete); |
1895 | 0 | } else if (parser->flags & F_CHUNKED) { |
1896 | | /* chunked encoding - ignore Content-Length header, |
1897 | | * prepare for a chunk */ |
1898 | 0 | UPDATE_STATE(s_chunk_size_start); |
1899 | 0 | } else if (parser->uses_transfer_encoding == 1) { |
1900 | 0 | if (parser->type == HTTP_REQUEST && !lenient) { |
1901 | | /* RFC 7230 3.3.3 */ |
1902 | | |
1903 | | /* If a Transfer-Encoding header field |
1904 | | * is present in a request and the chunked transfer coding is not |
1905 | | * the final encoding, the message body length cannot be determined |
1906 | | * reliably; the server MUST respond with the 400 (Bad Request) |
1907 | | * status code and then close the connection. |
1908 | | */ |
1909 | 0 | SET_ERRNO(HPE_INVALID_TRANSFER_ENCODING); |
1910 | 0 | RETURN(p - data); /* Error */ |
1911 | 0 | } else { |
1912 | | /* RFC 7230 3.3.3 */ |
1913 | | |
1914 | | /* If a Transfer-Encoding header field is present in a response and |
1915 | | * the chunked transfer coding is not the final encoding, the |
1916 | | * message body length is determined by reading the connection until |
1917 | | * it is closed by the server. |
1918 | | */ |
1919 | 0 | UPDATE_STATE(s_body_identity_eof); |
1920 | 0 | } |
1921 | 0 | } else { |
1922 | 0 | if (parser->content_length == 0) { |
1923 | | /* Content-Length header given but zero: Content-Length: 0\r\n */ |
1924 | 0 | UPDATE_STATE(NEW_MESSAGE()); |
1925 | 0 | CALLBACK_NOTIFY(message_complete); |
1926 | 0 | } else if (parser->content_length != ULLONG_MAX) { |
1927 | | /* Content-Length header given and non-zero */ |
1928 | 0 | UPDATE_STATE(s_body_identity); |
1929 | 0 | } else { |
1930 | 0 | if (!http_message_needs_eof(parser)) { |
1931 | | /* Assume content-length 0 - read the next */ |
1932 | 0 | UPDATE_STATE(NEW_MESSAGE()); |
1933 | 0 | CALLBACK_NOTIFY(message_complete); |
1934 | 0 | } else { |
1935 | | /* Read body until EOF */ |
1936 | 0 | UPDATE_STATE(s_body_identity_eof); |
1937 | 0 | } |
1938 | 0 | } |
1939 | 0 | } |
1940 | | |
1941 | 0 | break; |
1942 | 0 | } |
1943 | | |
1944 | 0 | case s_body_identity: |
1945 | 0 | { |
1946 | 0 | uint64_t to_read = MIN(parser->content_length, |
1947 | 0 | (uint64_t) ((data + len) - p)); |
1948 | |
|
1949 | 0 | assert(parser->content_length != 0 |
1950 | 0 | && parser->content_length != ULLONG_MAX); |
1951 | | |
1952 | | /* The difference between advancing content_length and p is because |
1953 | | * the latter will automaticaly advance on the next loop iteration. |
1954 | | * Further, if content_length ends up at 0, we want to see the last |
1955 | | * byte again for our message complete callback. |
1956 | | */ |
1957 | 0 | MARK(body); |
1958 | 0 | parser->content_length -= to_read; |
1959 | 0 | p += to_read - 1; |
1960 | |
|
1961 | 0 | if (parser->content_length == 0) { |
1962 | 0 | UPDATE_STATE(s_message_done); |
1963 | | |
1964 | | /* Mimic CALLBACK_DATA_NOADVANCE() but with one extra byte. |
1965 | | * |
1966 | | * The alternative to doing this is to wait for the next byte to |
1967 | | * trigger the data callback, just as in every other case. The |
1968 | | * problem with this is that this makes it difficult for the test |
1969 | | * harness to distinguish between complete-on-EOF and |
1970 | | * complete-on-length. It's not clear that this distinction is |
1971 | | * important for applications, but let's keep it for now. |
1972 | | */ |
1973 | 0 | CALLBACK_DATA_(body, p - body_mark + 1, p - data); |
1974 | 0 | REEXECUTE(); |
1975 | 0 | } |
1976 | | |
1977 | 0 | break; |
1978 | 0 | } |
1979 | | |
1980 | | /* read until EOF */ |
1981 | 0 | case s_body_identity_eof: |
1982 | 0 | MARK(body); |
1983 | 0 | p = data + len - 1; |
1984 | |
|
1985 | 0 | break; |
1986 | | |
1987 | 0 | case s_message_done: |
1988 | 0 | UPDATE_STATE(NEW_MESSAGE()); |
1989 | 0 | CALLBACK_NOTIFY(message_complete); |
1990 | 0 | if (parser->upgrade) { |
1991 | | /* Exit, the rest of the message is in a different protocol. */ |
1992 | 0 | RETURN((p - data) + 1); |
1993 | 0 | } |
1994 | 0 | break; |
1995 | | |
1996 | 0 | case s_chunk_size_start: |
1997 | 0 | { |
1998 | 0 | assert(nread == 1); |
1999 | 0 | assert(parser->flags & F_CHUNKED); |
2000 | |
|
2001 | 0 | unhex_val = unhex[(unsigned char)ch]; |
2002 | 0 | if (UNLIKELY(unhex_val == -1)) { |
2003 | 0 | SET_ERRNO(HPE_INVALID_CHUNK_SIZE); |
2004 | 0 | goto error; |
2005 | 0 | } |
2006 | | |
2007 | 0 | parser->content_length = unhex_val; |
2008 | 0 | UPDATE_STATE(s_chunk_size); |
2009 | 0 | break; |
2010 | 0 | } |
2011 | | |
2012 | 0 | case s_chunk_size: |
2013 | 0 | { |
2014 | 0 | uint64_t t; |
2015 | |
|
2016 | 0 | assert(parser->flags & F_CHUNKED); |
2017 | |
|
2018 | 0 | if (ch == CR) { |
2019 | 0 | UPDATE_STATE(s_chunk_size_almost_done); |
2020 | 0 | break; |
2021 | 0 | } |
2022 | | |
2023 | 0 | unhex_val = unhex[(unsigned char)ch]; |
2024 | |
|
2025 | 0 | if (unhex_val == -1) { |
2026 | 0 | if (ch == ';' || ch == ' ') { |
2027 | 0 | UPDATE_STATE(s_chunk_parameters); |
2028 | 0 | break; |
2029 | 0 | } |
2030 | | |
2031 | 0 | SET_ERRNO(HPE_INVALID_CHUNK_SIZE); |
2032 | 0 | goto error; |
2033 | 0 | } |
2034 | | |
2035 | 0 | t = parser->content_length; |
2036 | 0 | t *= 16; |
2037 | 0 | t += unhex_val; |
2038 | | |
2039 | | /* Overflow? Test against a conservative limit for simplicity. */ |
2040 | 0 | if (UNLIKELY((ULLONG_MAX - 16) / 16 < parser->content_length)) { |
2041 | 0 | SET_ERRNO(HPE_INVALID_CONTENT_LENGTH); |
2042 | 0 | goto error; |
2043 | 0 | } |
2044 | | |
2045 | 0 | parser->content_length = t; |
2046 | 0 | break; |
2047 | 0 | } |
2048 | | |
2049 | 0 | case s_chunk_parameters: |
2050 | 0 | { |
2051 | 0 | assert(parser->flags & F_CHUNKED); |
2052 | | /* just ignore this shit. TODO check for overflow */ |
2053 | 0 | if (ch == CR) { |
2054 | 0 | UPDATE_STATE(s_chunk_size_almost_done); |
2055 | 0 | break; |
2056 | 0 | } |
2057 | 0 | break; |
2058 | 0 | } |
2059 | | |
2060 | 0 | case s_chunk_size_almost_done: |
2061 | 0 | { |
2062 | 0 | assert(parser->flags & F_CHUNKED); |
2063 | 0 | STRICT_CHECK(ch != LF); |
2064 | |
|
2065 | 0 | parser->nread = 0; |
2066 | 0 | nread = 0; |
2067 | |
|
2068 | 0 | if (parser->content_length == 0) { |
2069 | 0 | parser->flags |= F_TRAILING; |
2070 | 0 | UPDATE_STATE(s_header_field_start); |
2071 | 0 | } else { |
2072 | 0 | UPDATE_STATE(s_chunk_data); |
2073 | 0 | } |
2074 | 0 | CALLBACK_NOTIFY(chunk_header); |
2075 | 0 | break; |
2076 | 0 | } |
2077 | | |
2078 | 0 | case s_chunk_data: |
2079 | 0 | { |
2080 | 0 | uint64_t to_read = MIN(parser->content_length, |
2081 | 0 | (uint64_t) ((data + len) - p)); |
2082 | |
|
2083 | 0 | assert(parser->flags & F_CHUNKED); |
2084 | 0 | assert(parser->content_length != 0 |
2085 | 0 | && parser->content_length != ULLONG_MAX); |
2086 | | |
2087 | | /* See the explanation in s_body_identity for why the content |
2088 | | * length and data pointers are managed this way. |
2089 | | */ |
2090 | 0 | MARK(body); |
2091 | 0 | parser->content_length -= to_read; |
2092 | 0 | p += to_read - 1; |
2093 | |
|
2094 | 0 | if (parser->content_length == 0) { |
2095 | 0 | UPDATE_STATE(s_chunk_data_almost_done); |
2096 | 0 | } |
2097 | |
|
2098 | 0 | break; |
2099 | 0 | } |
2100 | | |
2101 | 0 | case s_chunk_data_almost_done: |
2102 | 0 | assert(parser->flags & F_CHUNKED); |
2103 | 0 | assert(parser->content_length == 0); |
2104 | 0 | STRICT_CHECK(ch != CR); |
2105 | 0 | UPDATE_STATE(s_chunk_data_done); |
2106 | 0 | CALLBACK_DATA(body); |
2107 | 0 | break; |
2108 | | |
2109 | 0 | case s_chunk_data_done: |
2110 | 0 | assert(parser->flags & F_CHUNKED); |
2111 | 0 | STRICT_CHECK(ch != LF); |
2112 | 0 | parser->nread = 0; |
2113 | 0 | nread = 0; |
2114 | 0 | UPDATE_STATE(s_chunk_size_start); |
2115 | 0 | CALLBACK_NOTIFY(chunk_complete); |
2116 | 0 | break; |
2117 | | |
2118 | 0 | default: |
2119 | 0 | assert(0 && "unhandled state"); |
2120 | 0 | SET_ERRNO(HPE_INVALID_INTERNAL_STATE); |
2121 | 0 | goto error; |
2122 | 0 | } |
2123 | 0 | } |
2124 | | |
2125 | | /* Run callbacks for any marks that we have leftover after we ran out of |
2126 | | * bytes. There should be at most one of these set, so it's OK to invoke |
2127 | | * them in series (unset marks will not result in callbacks). |
2128 | | * |
2129 | | * We use the NOADVANCE() variety of callbacks here because 'p' has already |
2130 | | * overflowed 'data' and this allows us to correct for the off-by-one that |
2131 | | * we'd otherwise have (since CALLBACK_DATA() is meant to be run with a 'p' |
2132 | | * value that's in-bounds). |
2133 | | */ |
2134 | | |
2135 | 0 | assert(((header_field_mark ? 1 : 0) + |
2136 | 0 | (header_value_mark ? 1 : 0) + |
2137 | 0 | (url_mark ? 1 : 0) + |
2138 | 0 | (body_mark ? 1 : 0) + |
2139 | 0 | (status_mark ? 1 : 0)) <= 1); |
2140 | |
|
2141 | 0 | CALLBACK_DATA_NOADVANCE(header_field); |
2142 | 0 | CALLBACK_DATA_NOADVANCE(header_value); |
2143 | 0 | CALLBACK_DATA_NOADVANCE(url); |
2144 | 0 | CALLBACK_DATA_NOADVANCE(body); |
2145 | 0 | CALLBACK_DATA_NOADVANCE(status); |
2146 | | |
2147 | 0 | RETURN(len); |
2148 | |
|
2149 | 0 | error: |
2150 | 0 | if (HTTP_PARSER_ERRNO(parser) == HPE_OK) { |
2151 | 0 | SET_ERRNO(HPE_UNKNOWN); |
2152 | 0 | } |
2153 | |
|
2154 | 0 | RETURN(p - data); |
2155 | 0 | } |
2156 | | |
2157 | | |
2158 | | /* Does the parser need to see an EOF to find the end of the message? */ |
2159 | | int |
2160 | | http_message_needs_eof (const http_parser *parser) |
2161 | 0 | { |
2162 | 0 | if (parser->type == HTTP_REQUEST) { |
2163 | 0 | return 0; |
2164 | 0 | } |
2165 | | |
2166 | | /* See RFC 2616 section 4.4 */ |
2167 | 0 | if (parser->status_code / 100 == 1 || /* 1xx e.g. Continue */ |
2168 | 0 | parser->status_code == 204 || /* No Content */ |
2169 | 0 | parser->status_code == 304 || /* Not Modified */ |
2170 | 0 | parser->flags & F_SKIPBODY) { /* response to a HEAD request */ |
2171 | 0 | return 0; |
2172 | 0 | } |
2173 | | |
2174 | | /* RFC 7230 3.3.3, see `s_headers_almost_done` */ |
2175 | 0 | if ((parser->uses_transfer_encoding == 1) && |
2176 | 0 | (parser->flags & F_CHUNKED) == 0) { |
2177 | 0 | return 1; |
2178 | 0 | } |
2179 | | |
2180 | 0 | if ((parser->flags & F_CHUNKED) || parser->content_length != ULLONG_MAX) { |
2181 | 0 | return 0; |
2182 | 0 | } |
2183 | | |
2184 | 0 | return 1; |
2185 | 0 | } |
2186 | | |
2187 | | |
2188 | | int |
2189 | | http_should_keep_alive (const http_parser *parser) |
2190 | 0 | { |
2191 | 0 | if (parser->http_major > 0 && parser->http_minor > 0) { |
2192 | | /* HTTP/1.1 */ |
2193 | 0 | if (parser->flags & F_CONNECTION_CLOSE) { |
2194 | 0 | return 0; |
2195 | 0 | } |
2196 | 0 | } else { |
2197 | | /* HTTP/1.0 or earlier */ |
2198 | 0 | if (!(parser->flags & F_CONNECTION_KEEP_ALIVE)) { |
2199 | 0 | return 0; |
2200 | 0 | } |
2201 | 0 | } |
2202 | | |
2203 | 0 | return !http_message_needs_eof(parser); |
2204 | 0 | } |
2205 | | |
2206 | | |
2207 | | const char * |
2208 | | http_method_str (enum http_method m) |
2209 | 0 | { |
2210 | 0 | return ELEM_AT(method_strings, m, "<unknown>"); |
2211 | 0 | } |
2212 | | |
2213 | | const char * |
2214 | | http_status_str (enum http_status s) |
2215 | 0 | { |
2216 | 0 | switch (s) { |
2217 | 0 | #define XX(num, name, string) case HTTP_STATUS_##name: return #string; |
2218 | 0 | HTTP_STATUS_MAP(XX) |
2219 | 0 | #undef XX |
2220 | 0 | default: return "<unknown>"; |
2221 | 0 | } |
2222 | 0 | } |
2223 | | |
2224 | | void |
2225 | | http_parser_init (http_parser *parser, enum http_parser_type t) |
2226 | 0 | { |
2227 | 0 | void *data = parser->data; /* preserve application data */ |
2228 | 0 | memset(parser, 0, sizeof(*parser)); |
2229 | 0 | parser->data = data; |
2230 | 0 | parser->type = t; |
2231 | 0 | parser->state = (t == HTTP_REQUEST ? s_start_req : (t == HTTP_RESPONSE ? s_start_res : s_start_req_or_res)); |
2232 | 0 | parser->http_errno = HPE_OK; |
2233 | 0 | } |
2234 | | |
2235 | | void |
2236 | | http_parser_settings_init(http_parser_settings *settings) |
2237 | 0 | { |
2238 | 0 | memset(settings, 0, sizeof(*settings)); |
2239 | 0 | } |
2240 | | |
2241 | | const char * |
2242 | 0 | http_errno_name(enum http_errno err) { |
2243 | 0 | assert(((size_t) err) < ARRAY_SIZE(http_strerror_tab)); |
2244 | 0 | return http_strerror_tab[err].name; |
2245 | 0 | } |
2246 | | |
2247 | | const char * |
2248 | 0 | http_errno_description(enum http_errno err) { |
2249 | 0 | assert(((size_t) err) < ARRAY_SIZE(http_strerror_tab)); |
2250 | 0 | return http_strerror_tab[err].description; |
2251 | 0 | } |
2252 | | |
2253 | | static enum http_host_state |
2254 | 528k | http_parse_host_char(enum http_host_state s, const char ch) { |
2255 | 528k | switch(s) { |
2256 | 284k | case s_http_userinfo: |
2257 | 284k | case s_http_userinfo_start: |
2258 | 284k | if (ch == '@') { |
2259 | 206 | return s_http_host_start; |
2260 | 206 | } |
2261 | | |
2262 | 284k | if (IS_USERINFO_CHAR(ch)) { |
2263 | 284k | return s_http_userinfo; |
2264 | 284k | } |
2265 | 4 | break; |
2266 | | |
2267 | 566 | case s_http_host_start: |
2268 | 566 | if (ch == '[') { |
2269 | 182 | return s_http_host_v6_start; |
2270 | 182 | } |
2271 | | |
2272 | 384 | if (IS_HOST_CHAR(ch)) { |
2273 | 244 | return s_http_host; |
2274 | 244 | } |
2275 | | |
2276 | 140 | break; |
2277 | | |
2278 | 221k | case s_http_host: |
2279 | 221k | if (IS_HOST_CHAR(ch)) { |
2280 | 221k | return s_http_host; |
2281 | 221k | } |
2282 | | |
2283 | | /* fall through */ |
2284 | 150 | case s_http_host_v6_end: |
2285 | 150 | if (ch == ':') { |
2286 | 114 | return s_http_host_port_start; |
2287 | 114 | } |
2288 | | |
2289 | 36 | break; |
2290 | | |
2291 | 1.96k | case s_http_host_v6: |
2292 | 1.96k | if (ch == ']') { |
2293 | 9 | return s_http_host_v6_end; |
2294 | 9 | } |
2295 | | |
2296 | | /* fall through */ |
2297 | 2.13k | case s_http_host_v6_start: |
2298 | 2.13k | if (IS_HEX(ch) || ch == ':' || ch == '.') { |
2299 | 2.00k | return s_http_host_v6; |
2300 | 2.00k | } |
2301 | | |
2302 | 128 | if (s == s_http_host_v6 && ch == '%') { |
2303 | 92 | return s_http_host_v6_zone_start; |
2304 | 92 | } |
2305 | 36 | break; |
2306 | | |
2307 | 2.48k | case s_http_host_v6_zone: |
2308 | 2.48k | if (ch == ']') { |
2309 | 1 | return s_http_host_v6_end; |
2310 | 1 | } |
2311 | | |
2312 | | /* fall through */ |
2313 | 2.57k | case s_http_host_v6_zone_start: |
2314 | | /* RFC 6874 Zone ID consists of 1*( unreserved / pct-encoded) */ |
2315 | 2.57k | if (IS_ALPHANUM(ch) || ch == '%' || ch == '.' || ch == '-' || ch == '_' || |
2316 | 2.55k | ch == '~') { |
2317 | 2.55k | return s_http_host_v6_zone; |
2318 | 2.55k | } |
2319 | 18 | break; |
2320 | | |
2321 | 16.4k | case s_http_host_port: |
2322 | 16.5k | case s_http_host_port_start: |
2323 | 16.5k | if (IS_NUM(ch)) { |
2324 | 16.4k | return s_http_host_port; |
2325 | 16.4k | } |
2326 | | |
2327 | 11 | break; |
2328 | | |
2329 | 11 | default: |
2330 | 0 | break; |
2331 | 528k | } |
2332 | 245 | return s_http_host_dead; |
2333 | 528k | } |
2334 | | |
2335 | | static int |
2336 | 747 | http_parse_host(const char * buf, struct http_parser_url *u, int found_at) { |
2337 | 747 | enum http_host_state s; |
2338 | | |
2339 | 747 | const char *p; |
2340 | 747 | size_t buflen = u->field_data[UF_HOST].off + u->field_data[UF_HOST].len; |
2341 | | |
2342 | 747 | assert(u->field_set & (1 << UF_HOST)); |
2343 | | |
2344 | 747 | u->field_data[UF_HOST].len = 0; |
2345 | | |
2346 | 747 | s = found_at ? s_http_userinfo_start : s_http_host_start; |
2347 | | |
2348 | 528k | for (p = buf + u->field_data[UF_HOST].off; p < buf + buflen; p++) { |
2349 | 528k | enum http_host_state new_s = http_parse_host_char(s, *p); |
2350 | | |
2351 | 528k | if (new_s == s_http_host_dead) { |
2352 | 245 | return 1; |
2353 | 245 | } |
2354 | | |
2355 | 527k | switch(new_s) { |
2356 | 221k | case s_http_host: |
2357 | 221k | if (s != s_http_host) { |
2358 | 244 | u->field_data[UF_HOST].off = (uint16_t)(p - buf); |
2359 | 244 | } |
2360 | 221k | u->field_data[UF_HOST].len++; |
2361 | 221k | break; |
2362 | | |
2363 | 2.00k | case s_http_host_v6: |
2364 | 2.00k | if (s != s_http_host_v6) { |
2365 | 159 | u->field_data[UF_HOST].off = (uint16_t)(p - buf); |
2366 | 159 | } |
2367 | 2.00k | u->field_data[UF_HOST].len++; |
2368 | 2.00k | break; |
2369 | | |
2370 | 92 | case s_http_host_v6_zone_start: |
2371 | 2.64k | case s_http_host_v6_zone: |
2372 | 2.64k | u->field_data[UF_HOST].len++; |
2373 | 2.64k | break; |
2374 | | |
2375 | 16.4k | case s_http_host_port: |
2376 | 16.4k | if (s != s_http_host_port) { |
2377 | 69 | u->field_data[UF_PORT].off = (uint16_t)(p - buf); |
2378 | 69 | u->field_data[UF_PORT].len = 0; |
2379 | 69 | u->field_set |= (1 << UF_PORT); |
2380 | 69 | } |
2381 | 16.4k | u->field_data[UF_PORT].len++; |
2382 | 16.4k | break; |
2383 | | |
2384 | 284k | case s_http_userinfo: |
2385 | 284k | if (s != s_http_userinfo) { |
2386 | 191 | u->field_data[UF_USERINFO].off = (uint16_t)(p - buf); |
2387 | 191 | u->field_data[UF_USERINFO].len = 0; |
2388 | 191 | u->field_set |= (1 << UF_USERINFO); |
2389 | 191 | } |
2390 | 284k | u->field_data[UF_USERINFO].len++; |
2391 | 284k | break; |
2392 | | |
2393 | 512 | default: |
2394 | 512 | break; |
2395 | 527k | } |
2396 | 527k | s = new_s; |
2397 | 527k | } |
2398 | | |
2399 | | /* Make sure we don't end somewhere unexpected */ |
2400 | 502 | switch (s) { |
2401 | 175 | case s_http_host_start: |
2402 | 176 | case s_http_host_v6_start: |
2403 | 220 | case s_http_host_v6: |
2404 | 221 | case s_http_host_v6_zone_start: |
2405 | 293 | case s_http_host_v6_zone: |
2406 | 327 | case s_http_host_port_start: |
2407 | 327 | case s_http_userinfo: |
2408 | 329 | case s_http_userinfo_start: |
2409 | 329 | return 1; |
2410 | 173 | default: |
2411 | 173 | break; |
2412 | 502 | } |
2413 | | |
2414 | 173 | return 0; |
2415 | 502 | } |
2416 | | |
2417 | | void |
2418 | 966 | http_parser_url_init(struct http_parser_url *u) { |
2419 | 966 | memset(u, 0, sizeof(*u)); |
2420 | 966 | } |
2421 | | |
2422 | | int |
2423 | | http_parser_parse_url(const char *buf, size_t buflen, int is_connect, |
2424 | | struct http_parser_url *u) |
2425 | 1.93k | { |
2426 | 1.93k | enum state s; |
2427 | 1.93k | const char *p; |
2428 | 1.93k | enum http_parser_url_fields uf, old_uf; |
2429 | 1.93k | int found_at = 0; |
2430 | | |
2431 | 1.93k | if (buflen == 0) { |
2432 | 0 | return 1; |
2433 | 0 | } |
2434 | | |
2435 | 1.93k | u->port = u->field_set = 0; |
2436 | 1.93k | s = is_connect ? s_req_server_start : s_req_spaces_before_url; |
2437 | 1.93k | old_uf = UF_MAX; |
2438 | | |
2439 | 2.03M | for (p = buf; p < buf + buflen; p++) { |
2440 | 2.03M | s = parse_url_char(s, *p); |
2441 | | |
2442 | | /* Figure out the next field that we're operating on */ |
2443 | 2.03M | switch (s) { |
2444 | 952 | case s_dead: |
2445 | 952 | return 1; |
2446 | | |
2447 | | /* Skip delimeters */ |
2448 | 81 | case s_req_schema_slash: |
2449 | 118 | case s_req_schema_slash_slash: |
2450 | 145 | case s_req_server_start: |
2451 | 278 | case s_req_query_string_start: |
2452 | 780 | case s_req_fragment_start: |
2453 | 780 | continue; |
2454 | | |
2455 | 284k | case s_req_schema: |
2456 | 284k | uf = UF_SCHEMA; |
2457 | 284k | break; |
2458 | | |
2459 | 450 | case s_req_server_with_at: |
2460 | 450 | found_at = 1; |
2461 | | |
2462 | | /* fall through */ |
2463 | 905k | case s_req_server: |
2464 | 905k | uf = UF_HOST; |
2465 | 905k | break; |
2466 | | |
2467 | 404k | case s_req_path: |
2468 | 404k | uf = UF_PATH; |
2469 | 404k | break; |
2470 | | |
2471 | 435k | case s_req_query_string: |
2472 | 435k | uf = UF_QUERY; |
2473 | 435k | break; |
2474 | | |
2475 | 862 | case s_req_fragment: |
2476 | 862 | uf = UF_FRAGMENT; |
2477 | 862 | break; |
2478 | | |
2479 | 0 | default: |
2480 | 0 | assert(!"Unexpected state"); |
2481 | 0 | return 1; |
2482 | 2.03M | } |
2483 | | |
2484 | | /* Nothing's changed; soldier on */ |
2485 | 2.03M | if (uf == old_uf) { |
2486 | 2.03M | u->field_data[uf].len++; |
2487 | 2.03M | continue; |
2488 | 2.03M | } |
2489 | | |
2490 | 1.40k | u->field_data[uf].off = (uint16_t)(p - buf); |
2491 | 1.40k | u->field_data[uf].len = 1; |
2492 | | |
2493 | 1.40k | u->field_set |= (1 << uf); |
2494 | 1.40k | old_uf = uf; |
2495 | 1.40k | } |
2496 | | |
2497 | | /* host must be present if there is a schema */ |
2498 | | /* parsing http:///toto will fail */ |
2499 | 980 | if ((u->field_set & (1 << UF_SCHEMA)) && |
2500 | 50 | (u->field_set & (1 << UF_HOST)) == 0) { |
2501 | 25 | return 1; |
2502 | 25 | } |
2503 | | |
2504 | 955 | if (u->field_set & (1 << UF_HOST)) { |
2505 | 747 | if (http_parse_host(buf, u, found_at) != 0) { |
2506 | 574 | return 1; |
2507 | 574 | } |
2508 | 747 | } |
2509 | | |
2510 | | /* CONNECT requests can only contain "hostname:port" */ |
2511 | 381 | if (is_connect && u->field_set != ((1 << UF_HOST)|(1 << UF_PORT))) { |
2512 | 200 | return 1; |
2513 | 200 | } |
2514 | | |
2515 | 181 | if (u->field_set & (1 << UF_PORT)) { |
2516 | 69 | uint16_t off; |
2517 | 69 | uint16_t len; |
2518 | 69 | const char* p; |
2519 | 69 | const char* end; |
2520 | 69 | unsigned long v; |
2521 | | |
2522 | 69 | off = u->field_data[UF_PORT].off; |
2523 | 69 | len = u->field_data[UF_PORT].len; |
2524 | 69 | end = buf + off + len; |
2525 | | |
2526 | | /* NOTE: The characters are already validated and are in the [0-9] range */ |
2527 | 69 | assert((size_t) (off + len) <= buflen && "Port number overflow"); |
2528 | 69 | v = 0; |
2529 | 578 | for (p = buf + off; p < end; p++) { |
2530 | 522 | v *= 10; |
2531 | 522 | v += *p - '0'; |
2532 | | |
2533 | | /* Ports have a max value of 2^16 */ |
2534 | 522 | if (v > 0xffff) { |
2535 | 13 | return 1; |
2536 | 13 | } |
2537 | 522 | } |
2538 | | |
2539 | 56 | u->port = (uint16_t) v; |
2540 | 56 | } |
2541 | | |
2542 | 168 | return 0; |
2543 | 181 | } |
2544 | | |
2545 | | void |
2546 | 0 | http_parser_pause(http_parser *parser, int paused) { |
2547 | | /* Users should only be pausing/unpausing a parser that is not in an error |
2548 | | * state. In non-debug builds, there's not much that we can do about this |
2549 | | * other than ignore it. |
2550 | | */ |
2551 | 0 | if (HTTP_PARSER_ERRNO(parser) == HPE_OK || |
2552 | 0 | HTTP_PARSER_ERRNO(parser) == HPE_PAUSED) { |
2553 | 0 | uint32_t nread = parser->nread; /* used by the SET_ERRNO macro */ |
2554 | 0 | SET_ERRNO((paused) ? HPE_PAUSED : HPE_OK); |
2555 | 0 | } else { |
2556 | 0 | assert(0 && "Attempting to pause parser in error state"); |
2557 | 0 | } |
2558 | 0 | } |
2559 | | |
2560 | | int |
2561 | 0 | http_body_is_final(const struct http_parser *parser) { |
2562 | 0 | return parser->state == s_message_done; |
2563 | 0 | } |
2564 | | |
2565 | | unsigned long |
2566 | 0 | http_parser_version(void) { |
2567 | 0 | return HTTP_PARSER_VERSION_MAJOR * 0x10000 | |
2568 | 0 | HTTP_PARSER_VERSION_MINOR * 0x00100 | |
2569 | 0 | HTTP_PARSER_VERSION_PATCH * 0x00001; |
2570 | 0 | } |
2571 | | |
2572 | | void |
2573 | 0 | http_parser_set_max_header_size(uint32_t size) { |
2574 | 0 | max_header_size = size; |
2575 | 0 | } |