giraffesoft-unicorn 0.93.5
Sign up to get free protection for your applications and to get access to all the features.
- data/.CHANGELOG.old +25 -0
- data/.document +16 -0
- data/.gitignore +20 -0
- data/.mailmap +26 -0
- data/CONTRIBUTORS +31 -0
- data/COPYING +339 -0
- data/DESIGN +105 -0
- data/Documentation/.gitignore +5 -0
- data/Documentation/GNUmakefile +30 -0
- data/Documentation/unicorn.1.txt +167 -0
- data/Documentation/unicorn_rails.1.txt +169 -0
- data/GIT-VERSION-GEN +40 -0
- data/GNUmakefile +270 -0
- data/HACKING +113 -0
- data/KNOWN_ISSUES +40 -0
- data/LICENSE +55 -0
- data/PHILOSOPHY +144 -0
- data/README +153 -0
- data/Rakefile +108 -0
- data/SIGNALS +97 -0
- data/TODO +16 -0
- data/TUNING +70 -0
- data/bin/unicorn +165 -0
- data/bin/unicorn_rails +208 -0
- data/examples/echo.ru +27 -0
- data/examples/git.ru +13 -0
- data/examples/init.sh +53 -0
- data/ext/unicorn_http/c_util.h +107 -0
- data/ext/unicorn_http/common_field_optimization.h +111 -0
- data/ext/unicorn_http/ext_help.h +73 -0
- data/ext/unicorn_http/extconf.rb +14 -0
- data/ext/unicorn_http/global_variables.h +91 -0
- data/ext/unicorn_http/unicorn_http.rl +715 -0
- data/ext/unicorn_http/unicorn_http_common.rl +74 -0
- data/lib/unicorn.rb +730 -0
- data/lib/unicorn/app/exec_cgi.rb +150 -0
- data/lib/unicorn/app/inetd.rb +109 -0
- data/lib/unicorn/app/old_rails.rb +31 -0
- data/lib/unicorn/app/old_rails/static.rb +60 -0
- data/lib/unicorn/cgi_wrapper.rb +145 -0
- data/lib/unicorn/configurator.rb +403 -0
- data/lib/unicorn/const.rb +37 -0
- data/lib/unicorn/http_request.rb +74 -0
- data/lib/unicorn/http_response.rb +74 -0
- data/lib/unicorn/launcher.rb +39 -0
- data/lib/unicorn/socket_helper.rb +138 -0
- data/lib/unicorn/tee_input.rb +174 -0
- data/lib/unicorn/util.rb +64 -0
- data/local.mk.sample +53 -0
- data/setup.rb +1586 -0
- data/test/aggregate.rb +15 -0
- data/test/benchmark/README +50 -0
- data/test/benchmark/dd.ru +18 -0
- data/test/exec/README +5 -0
- data/test/exec/test_exec.rb +855 -0
- data/test/rails/app-1.2.3/.gitignore +2 -0
- data/test/rails/app-1.2.3/Rakefile +7 -0
- data/test/rails/app-1.2.3/app/controllers/application.rb +6 -0
- data/test/rails/app-1.2.3/app/controllers/foo_controller.rb +36 -0
- data/test/rails/app-1.2.3/app/helpers/application_helper.rb +4 -0
- data/test/rails/app-1.2.3/config/boot.rb +11 -0
- data/test/rails/app-1.2.3/config/database.yml +12 -0
- data/test/rails/app-1.2.3/config/environment.rb +13 -0
- data/test/rails/app-1.2.3/config/environments/development.rb +9 -0
- data/test/rails/app-1.2.3/config/environments/production.rb +5 -0
- data/test/rails/app-1.2.3/config/routes.rb +6 -0
- data/test/rails/app-1.2.3/db/.gitignore +0 -0
- data/test/rails/app-1.2.3/public/404.html +1 -0
- data/test/rails/app-1.2.3/public/500.html +1 -0
- data/test/rails/app-2.0.2/.gitignore +2 -0
- data/test/rails/app-2.0.2/Rakefile +7 -0
- data/test/rails/app-2.0.2/app/controllers/application.rb +4 -0
- data/test/rails/app-2.0.2/app/controllers/foo_controller.rb +36 -0
- data/test/rails/app-2.0.2/app/helpers/application_helper.rb +4 -0
- data/test/rails/app-2.0.2/config/boot.rb +11 -0
- data/test/rails/app-2.0.2/config/database.yml +12 -0
- data/test/rails/app-2.0.2/config/environment.rb +17 -0
- data/test/rails/app-2.0.2/config/environments/development.rb +8 -0
- data/test/rails/app-2.0.2/config/environments/production.rb +5 -0
- data/test/rails/app-2.0.2/config/routes.rb +6 -0
- data/test/rails/app-2.0.2/db/.gitignore +0 -0
- data/test/rails/app-2.0.2/public/404.html +1 -0
- data/test/rails/app-2.0.2/public/500.html +1 -0
- data/test/rails/app-2.1.2/.gitignore +2 -0
- data/test/rails/app-2.1.2/Rakefile +7 -0
- data/test/rails/app-2.1.2/app/controllers/application.rb +4 -0
- data/test/rails/app-2.1.2/app/controllers/foo_controller.rb +36 -0
- data/test/rails/app-2.1.2/app/helpers/application_helper.rb +4 -0
- data/test/rails/app-2.1.2/config/boot.rb +111 -0
- data/test/rails/app-2.1.2/config/database.yml +12 -0
- data/test/rails/app-2.1.2/config/environment.rb +17 -0
- data/test/rails/app-2.1.2/config/environments/development.rb +7 -0
- data/test/rails/app-2.1.2/config/environments/production.rb +5 -0
- data/test/rails/app-2.1.2/config/routes.rb +6 -0
- data/test/rails/app-2.1.2/db/.gitignore +0 -0
- data/test/rails/app-2.1.2/public/404.html +1 -0
- data/test/rails/app-2.1.2/public/500.html +1 -0
- data/test/rails/app-2.2.2/.gitignore +2 -0
- data/test/rails/app-2.2.2/Rakefile +7 -0
- data/test/rails/app-2.2.2/app/controllers/application.rb +4 -0
- data/test/rails/app-2.2.2/app/controllers/foo_controller.rb +36 -0
- data/test/rails/app-2.2.2/app/helpers/application_helper.rb +4 -0
- data/test/rails/app-2.2.2/config/boot.rb +111 -0
- data/test/rails/app-2.2.2/config/database.yml +12 -0
- data/test/rails/app-2.2.2/config/environment.rb +17 -0
- data/test/rails/app-2.2.2/config/environments/development.rb +7 -0
- data/test/rails/app-2.2.2/config/environments/production.rb +5 -0
- data/test/rails/app-2.2.2/config/routes.rb +6 -0
- data/test/rails/app-2.2.2/db/.gitignore +0 -0
- data/test/rails/app-2.2.2/public/404.html +1 -0
- data/test/rails/app-2.2.2/public/500.html +1 -0
- data/test/rails/app-2.3.3.1/.gitignore +2 -0
- data/test/rails/app-2.3.3.1/Rakefile +7 -0
- data/test/rails/app-2.3.3.1/app/controllers/application_controller.rb +5 -0
- data/test/rails/app-2.3.3.1/app/controllers/foo_controller.rb +36 -0
- data/test/rails/app-2.3.3.1/app/helpers/application_helper.rb +4 -0
- data/test/rails/app-2.3.3.1/config/boot.rb +109 -0
- data/test/rails/app-2.3.3.1/config/database.yml +12 -0
- data/test/rails/app-2.3.3.1/config/environment.rb +17 -0
- data/test/rails/app-2.3.3.1/config/environments/development.rb +7 -0
- data/test/rails/app-2.3.3.1/config/environments/production.rb +6 -0
- data/test/rails/app-2.3.3.1/config/routes.rb +6 -0
- data/test/rails/app-2.3.3.1/db/.gitignore +0 -0
- data/test/rails/app-2.3.3.1/public/404.html +1 -0
- data/test/rails/app-2.3.3.1/public/500.html +1 -0
- data/test/rails/app-2.3.3.1/public/x.txt +1 -0
- data/test/rails/test_rails.rb +280 -0
- data/test/test_helper.rb +296 -0
- data/test/unit/test_configurator.rb +150 -0
- data/test/unit/test_http_parser.rb +492 -0
- data/test/unit/test_http_parser_ng.rb +308 -0
- data/test/unit/test_request.rb +184 -0
- data/test/unit/test_response.rb +110 -0
- data/test/unit/test_server.rb +188 -0
- data/test/unit/test_signals.rb +202 -0
- data/test/unit/test_socket_helper.rb +133 -0
- data/test/unit/test_tee_input.rb +229 -0
- data/test/unit/test_upload.rb +297 -0
- data/test/unit/test_util.rb +96 -0
- data/unicorn.gemspec +42 -0
- metadata +228 -0
@@ -0,0 +1,715 @@
|
|
1
|
+
/**
|
2
|
+
* Copyright (c) 2009 Eric Wong (all bugs are Eric's fault)
|
3
|
+
* Copyright (c) 2005 Zed A. Shaw
|
4
|
+
* You can redistribute it and/or modify it under the same terms as Ruby.
|
5
|
+
*/
|
6
|
+
#include "ruby.h"
|
7
|
+
#include "ext_help.h"
|
8
|
+
#include <assert.h>
|
9
|
+
#include <string.h>
|
10
|
+
#include <sys/types.h>
|
11
|
+
#include "common_field_optimization.h"
|
12
|
+
#include "global_variables.h"
|
13
|
+
#include "c_util.h"
|
14
|
+
|
15
|
+
#define UH_FL_CHUNKED 0x1
|
16
|
+
#define UH_FL_HASBODY 0x2
|
17
|
+
#define UH_FL_INBODY 0x4
|
18
|
+
#define UH_FL_HASTRAILER 0x8
|
19
|
+
#define UH_FL_INTRAILER 0x10
|
20
|
+
#define UH_FL_INCHUNK 0x20
|
21
|
+
#define UH_FL_KAMETHOD 0x40
|
22
|
+
#define UH_FL_KAVERSION 0x80
|
23
|
+
#define UH_FL_HASHEADER 0x100
|
24
|
+
|
25
|
+
/* both of these flags need to be set for keepalive to be supported */
|
26
|
+
#define UH_FL_KEEPALIVE (UH_FL_KAMETHOD | UH_FL_KAVERSION)
|
27
|
+
|
28
|
+
struct http_parser {
|
29
|
+
int cs; /* Ragel internal state */
|
30
|
+
unsigned int flags;
|
31
|
+
size_t mark;
|
32
|
+
union { /* these 3 fields don't nest */
|
33
|
+
size_t field;
|
34
|
+
size_t query;
|
35
|
+
size_t offset;
|
36
|
+
} start;
|
37
|
+
union {
|
38
|
+
size_t field_len; /* only used during header processing */
|
39
|
+
size_t dest_offset; /* only used during body processing */
|
40
|
+
} s;
|
41
|
+
VALUE cont; /* Qfalse: unset, Qnil: ignored header, T_STRING: append */
|
42
|
+
union {
|
43
|
+
off_t content;
|
44
|
+
off_t chunk;
|
45
|
+
} len;
|
46
|
+
};
|
47
|
+
|
48
|
+
static void finalize_header(struct http_parser *hp, VALUE req);
|
49
|
+
|
50
|
+
#define REMAINING (unsigned long)(pe - p)
|
51
|
+
#define LEN(AT, FPC) (FPC - buffer - hp->AT)
|
52
|
+
#define MARK(M,FPC) (hp->M = (FPC) - buffer)
|
53
|
+
#define PTR_TO(F) (buffer + hp->F)
|
54
|
+
#define STR_NEW(M,FPC) rb_str_new(PTR_TO(M), LEN(M, FPC))
|
55
|
+
|
56
|
+
#define HP_FL_TEST(hp,fl) ((hp)->flags & (UH_FL_##fl))
|
57
|
+
#define HP_FL_SET(hp,fl) ((hp)->flags |= (UH_FL_##fl))
|
58
|
+
#define HP_FL_UNSET(hp,fl) ((hp)->flags &= ~(UH_FL_##fl))
|
59
|
+
#define HP_FL_ALL(hp,fl) (HP_FL_TEST(hp, fl) == (UH_FL_##fl))
|
60
|
+
|
61
|
+
/*
|
62
|
+
* handles values of the "Connection:" header, keepalive is implied
|
63
|
+
* for HTTP/1.1 but needs to be explicitly enabled with HTTP/1.0
|
64
|
+
* Additionally, we require GET/HEAD requests to support keepalive.
|
65
|
+
*/
|
66
|
+
static void hp_keepalive_connection(struct http_parser *hp, VALUE val)
|
67
|
+
{
|
68
|
+
/* REQUEST_METHOD is always set before any headers */
|
69
|
+
if (HP_FL_TEST(hp, KAMETHOD)) {
|
70
|
+
if (STR_CSTR_CASE_EQ(val, "keep-alive")) {
|
71
|
+
/* basically have HTTP/1.0 masquerade as HTTP/1.1+ */
|
72
|
+
HP_FL_SET(hp, KAVERSION);
|
73
|
+
} else if (STR_CSTR_CASE_EQ(val, "close")) {
|
74
|
+
/*
|
75
|
+
* it doesn't matter what HTTP version or request method we have,
|
76
|
+
* if a client says "Connection: close", we disable keepalive
|
77
|
+
*/
|
78
|
+
HP_FL_UNSET(hp, KEEPALIVE);
|
79
|
+
} else {
|
80
|
+
/*
|
81
|
+
* client could've sent anything, ignore it for now. Maybe
|
82
|
+
* "HP_FL_UNSET(hp, KEEPALIVE);" just in case?
|
83
|
+
* Raising an exception might be too mean...
|
84
|
+
*/
|
85
|
+
}
|
86
|
+
}
|
87
|
+
}
|
88
|
+
|
89
|
+
static void
|
90
|
+
request_method(struct http_parser *hp, VALUE req, const char *ptr, size_t len)
|
91
|
+
{
|
92
|
+
VALUE v;
|
93
|
+
|
94
|
+
/*
|
95
|
+
* we only support keepalive for GET and HEAD requests for now other
|
96
|
+
* methods are too rarely seen to be worth optimizing. POST is unsafe
|
97
|
+
* since some clients send extra bytes after POST bodies.
|
98
|
+
*/
|
99
|
+
if (CONST_MEM_EQ("GET", ptr, len)) {
|
100
|
+
HP_FL_SET(hp, KAMETHOD);
|
101
|
+
v = g_GET;
|
102
|
+
} else if (CONST_MEM_EQ("HEAD", ptr, len)) {
|
103
|
+
HP_FL_SET(hp, KAMETHOD);
|
104
|
+
v = g_HEAD;
|
105
|
+
} else {
|
106
|
+
v = rb_str_new(ptr, len);
|
107
|
+
}
|
108
|
+
rb_hash_aset(req, g_request_method, v);
|
109
|
+
}
|
110
|
+
|
111
|
+
static void
|
112
|
+
http_version(struct http_parser *hp, VALUE req, const char *ptr, size_t len)
|
113
|
+
{
|
114
|
+
VALUE v;
|
115
|
+
|
116
|
+
HP_FL_SET(hp, HASHEADER);
|
117
|
+
|
118
|
+
if (CONST_MEM_EQ("HTTP/1.1", ptr, len)) {
|
119
|
+
/* HTTP/1.1 implies keepalive unless "Connection: close" is set */
|
120
|
+
HP_FL_SET(hp, KAVERSION);
|
121
|
+
v = g_http_11;
|
122
|
+
} else if (CONST_MEM_EQ("HTTP/1.0", ptr, len)) {
|
123
|
+
v = g_http_10;
|
124
|
+
} else {
|
125
|
+
v = rb_str_new(ptr, len);
|
126
|
+
}
|
127
|
+
rb_hash_aset(req, g_server_protocol, v);
|
128
|
+
rb_hash_aset(req, g_http_version, v);
|
129
|
+
}
|
130
|
+
|
131
|
+
static inline void hp_invalid_if_trailer(struct http_parser *hp)
|
132
|
+
{
|
133
|
+
if (HP_FL_TEST(hp, INTRAILER))
|
134
|
+
rb_raise(eHttpParserError, "invalid Trailer");
|
135
|
+
}
|
136
|
+
|
137
|
+
static void write_cont_value(struct http_parser *hp,
|
138
|
+
const char *buffer, const char *p)
|
139
|
+
{
|
140
|
+
char *vptr;
|
141
|
+
|
142
|
+
if (hp->cont == Qfalse)
|
143
|
+
rb_raise(eHttpParserError, "invalid continuation line");
|
144
|
+
if (NIL_P(hp->cont))
|
145
|
+
return; /* we're ignoring this header (probably Host:) */
|
146
|
+
|
147
|
+
assert(TYPE(hp->cont) == T_STRING && "continuation line is not a string");
|
148
|
+
assert(hp->mark > 0 && "impossible continuation line offset");
|
149
|
+
|
150
|
+
if (LEN(mark, p) == 0)
|
151
|
+
return;
|
152
|
+
|
153
|
+
if (RSTRING_LEN(hp->cont) > 0)
|
154
|
+
--hp->mark;
|
155
|
+
|
156
|
+
vptr = (char *)PTR_TO(mark);
|
157
|
+
|
158
|
+
if (RSTRING_LEN(hp->cont) > 0) {
|
159
|
+
assert((' ' == *vptr || '\t' == *vptr) && "invalid leading white space");
|
160
|
+
*vptr = ' ';
|
161
|
+
}
|
162
|
+
rb_str_buf_cat(hp->cont, vptr, LEN(mark, p));
|
163
|
+
}
|
164
|
+
|
165
|
+
static void write_value(VALUE req, struct http_parser *hp,
|
166
|
+
const char *buffer, const char *p)
|
167
|
+
{
|
168
|
+
VALUE f = find_common_field(PTR_TO(start.field), hp->s.field_len);
|
169
|
+
VALUE v;
|
170
|
+
VALUE e;
|
171
|
+
|
172
|
+
VALIDATE_MAX_LENGTH(LEN(mark, p), FIELD_VALUE);
|
173
|
+
v = LEN(mark, p) == 0 ? rb_str_buf_new(128) : STR_NEW(mark, p);
|
174
|
+
if (NIL_P(f)) {
|
175
|
+
VALIDATE_MAX_LENGTH(hp->s.field_len, FIELD_NAME);
|
176
|
+
f = uncommon_field(PTR_TO(start.field), hp->s.field_len);
|
177
|
+
} else if (f == g_http_connection) {
|
178
|
+
hp_keepalive_connection(hp, v);
|
179
|
+
} else if (f == g_content_length) {
|
180
|
+
hp->len.content = parse_length(RSTRING_PTR(v), RSTRING_LEN(v));
|
181
|
+
if (hp->len.content < 0)
|
182
|
+
rb_raise(eHttpParserError, "invalid Content-Length");
|
183
|
+
HP_FL_SET(hp, HASBODY);
|
184
|
+
hp_invalid_if_trailer(hp);
|
185
|
+
} else if (f == g_http_transfer_encoding) {
|
186
|
+
if (STR_CSTR_CASE_EQ(v, "chunked")) {
|
187
|
+
HP_FL_SET(hp, CHUNKED);
|
188
|
+
HP_FL_SET(hp, HASBODY);
|
189
|
+
}
|
190
|
+
hp_invalid_if_trailer(hp);
|
191
|
+
} else if (f == g_http_trailer) {
|
192
|
+
HP_FL_SET(hp, HASTRAILER);
|
193
|
+
hp_invalid_if_trailer(hp);
|
194
|
+
} else {
|
195
|
+
assert(TYPE(f) == T_STRING && "memoized object is not a string");
|
196
|
+
assert_frozen(f);
|
197
|
+
}
|
198
|
+
|
199
|
+
e = rb_hash_aref(req, f);
|
200
|
+
if (NIL_P(e)) {
|
201
|
+
hp->cont = rb_hash_aset(req, f, v);
|
202
|
+
} else if (f == g_http_host) {
|
203
|
+
/*
|
204
|
+
* ignored, absolute URLs in REQUEST_URI take precedence over
|
205
|
+
* the Host: header (ref: rfc 2616, section 5.2.1)
|
206
|
+
*/
|
207
|
+
hp->cont = Qnil;
|
208
|
+
} else {
|
209
|
+
rb_str_buf_cat(e, ",", 1);
|
210
|
+
hp->cont = rb_str_buf_append(e, v);
|
211
|
+
}
|
212
|
+
}
|
213
|
+
|
214
|
+
/** Machine **/
|
215
|
+
|
216
|
+
%%{
|
217
|
+
machine http_parser;
|
218
|
+
|
219
|
+
action mark {MARK(mark, fpc); }
|
220
|
+
|
221
|
+
action start_field { MARK(start.field, fpc); }
|
222
|
+
action snake_upcase_field { snake_upcase_char((char *)fpc); }
|
223
|
+
action downcase_char { downcase_char((char *)fpc); }
|
224
|
+
action write_field { hp->s.field_len = LEN(start.field, fpc); }
|
225
|
+
action start_value { MARK(mark, fpc); }
|
226
|
+
action write_value { write_value(req, hp, buffer, fpc); }
|
227
|
+
action write_cont_value { write_cont_value(hp, buffer, fpc); }
|
228
|
+
action request_method {
|
229
|
+
request_method(hp, req, PTR_TO(mark), LEN(mark, fpc));
|
230
|
+
}
|
231
|
+
action scheme {
|
232
|
+
rb_hash_aset(req, g_rack_url_scheme, STR_NEW(mark, fpc));
|
233
|
+
}
|
234
|
+
action host {
|
235
|
+
rb_hash_aset(req, g_http_host, STR_NEW(mark, fpc));
|
236
|
+
}
|
237
|
+
action request_uri {
|
238
|
+
size_t len = LEN(mark, fpc);
|
239
|
+
VALUE str;
|
240
|
+
|
241
|
+
VALIDATE_MAX_LENGTH(len, REQUEST_URI);
|
242
|
+
str = rb_hash_aset(req, g_request_uri, STR_NEW(mark, fpc));
|
243
|
+
/*
|
244
|
+
* "OPTIONS * HTTP/1.1\r\n" is a valid request, but we can't have '*'
|
245
|
+
* in REQUEST_PATH or PATH_INFO or else Rack::Lint will complain
|
246
|
+
*/
|
247
|
+
if (STR_CSTR_EQ(str, "*")) {
|
248
|
+
str = rb_str_new(NULL, 0);
|
249
|
+
rb_hash_aset(req, g_path_info, str);
|
250
|
+
rb_hash_aset(req, g_request_path, str);
|
251
|
+
}
|
252
|
+
}
|
253
|
+
action fragment {
|
254
|
+
VALIDATE_MAX_LENGTH(LEN(mark, fpc), FRAGMENT);
|
255
|
+
rb_hash_aset(req, g_fragment, STR_NEW(mark, fpc));
|
256
|
+
}
|
257
|
+
action start_query {MARK(start.query, fpc); }
|
258
|
+
action query_string {
|
259
|
+
VALIDATE_MAX_LENGTH(LEN(start.query, fpc), QUERY_STRING);
|
260
|
+
rb_hash_aset(req, g_query_string, STR_NEW(start.query, fpc));
|
261
|
+
}
|
262
|
+
action http_version { http_version(hp, req, PTR_TO(mark), LEN(mark, fpc)); }
|
263
|
+
action request_path {
|
264
|
+
VALUE val;
|
265
|
+
size_t len = LEN(mark, fpc);
|
266
|
+
|
267
|
+
VALIDATE_MAX_LENGTH(len, REQUEST_PATH);
|
268
|
+
val = rb_hash_aset(req, g_request_path, STR_NEW(mark, fpc));
|
269
|
+
|
270
|
+
/* rack says PATH_INFO must start with "/" or be empty */
|
271
|
+
if (!STR_CSTR_EQ(val, "*"))
|
272
|
+
rb_hash_aset(req, g_path_info, val);
|
273
|
+
}
|
274
|
+
action add_to_chunk_size {
|
275
|
+
hp->len.chunk = step_incr(hp->len.chunk, fc, 16);
|
276
|
+
if (hp->len.chunk < 0)
|
277
|
+
rb_raise(eHttpParserError, "invalid chunk size");
|
278
|
+
}
|
279
|
+
action header_done {
|
280
|
+
finalize_header(hp, req);
|
281
|
+
|
282
|
+
cs = http_parser_first_final;
|
283
|
+
if (HP_FL_TEST(hp, HASBODY)) {
|
284
|
+
HP_FL_SET(hp, INBODY);
|
285
|
+
if (HP_FL_TEST(hp, CHUNKED))
|
286
|
+
cs = http_parser_en_ChunkedBody;
|
287
|
+
} else {
|
288
|
+
assert(!HP_FL_TEST(hp, CHUNKED) && "chunked encoding without body!");
|
289
|
+
}
|
290
|
+
/*
|
291
|
+
* go back to Ruby so we can call the Rack application, we'll reenter
|
292
|
+
* the parser iff the body needs to be processed.
|
293
|
+
*/
|
294
|
+
goto post_exec;
|
295
|
+
}
|
296
|
+
|
297
|
+
action end_trailers {
|
298
|
+
cs = http_parser_first_final;
|
299
|
+
goto post_exec;
|
300
|
+
}
|
301
|
+
|
302
|
+
action end_chunked_body {
|
303
|
+
if (HP_FL_TEST(hp, HASTRAILER)) {
|
304
|
+
HP_FL_SET(hp, INTRAILER);
|
305
|
+
cs = http_parser_en_Trailers;
|
306
|
+
} else {
|
307
|
+
cs = http_parser_first_final;
|
308
|
+
}
|
309
|
+
++p;
|
310
|
+
assert(p <= pe && "buffer overflow after chunked body");
|
311
|
+
goto post_exec;
|
312
|
+
}
|
313
|
+
|
314
|
+
action skip_chunk_data {
|
315
|
+
skip_chunk_data_hack: {
|
316
|
+
size_t nr = MIN(hp->len.chunk, REMAINING);
|
317
|
+
memcpy(RSTRING_PTR(req) + hp->s.dest_offset, fpc, nr);
|
318
|
+
hp->s.dest_offset += nr;
|
319
|
+
hp->len.chunk -= nr;
|
320
|
+
p += nr;
|
321
|
+
assert(hp->len.chunk >= 0 && "negative chunk length");
|
322
|
+
if (hp->len.chunk > REMAINING) {
|
323
|
+
HP_FL_SET(hp, INCHUNK);
|
324
|
+
goto post_exec;
|
325
|
+
} else {
|
326
|
+
fhold;
|
327
|
+
fgoto chunk_end;
|
328
|
+
}
|
329
|
+
}}
|
330
|
+
|
331
|
+
include unicorn_http_common "unicorn_http_common.rl";
|
332
|
+
}%%
|
333
|
+
|
334
|
+
/** Data **/
|
335
|
+
%% write data;
|
336
|
+
|
337
|
+
static void http_parser_init(struct http_parser *hp)
|
338
|
+
{
|
339
|
+
int cs = 0;
|
340
|
+
memset(hp, 0, sizeof(struct http_parser));
|
341
|
+
hp->cont = Qfalse; /* zero on MRI, should be optimized away by above */
|
342
|
+
%% write init;
|
343
|
+
hp->cs = cs;
|
344
|
+
}
|
345
|
+
|
346
|
+
/** exec **/
|
347
|
+
static void http_parser_execute(struct http_parser *hp,
|
348
|
+
VALUE req, const char *buffer, size_t len)
|
349
|
+
{
|
350
|
+
const char *p, *pe;
|
351
|
+
int cs = hp->cs;
|
352
|
+
size_t off = hp->start.offset;
|
353
|
+
|
354
|
+
if (cs == http_parser_first_final)
|
355
|
+
return;
|
356
|
+
|
357
|
+
assert(off <= len && "offset past end of buffer");
|
358
|
+
|
359
|
+
p = buffer+off;
|
360
|
+
pe = buffer+len;
|
361
|
+
|
362
|
+
assert(pe - p == len - off && "pointers aren't same distance");
|
363
|
+
|
364
|
+
if (HP_FL_TEST(hp, INCHUNK)) {
|
365
|
+
HP_FL_UNSET(hp, INCHUNK);
|
366
|
+
goto skip_chunk_data_hack;
|
367
|
+
}
|
368
|
+
%% write exec;
|
369
|
+
post_exec: /* "_out:" also goes here */
|
370
|
+
if (hp->cs != http_parser_error)
|
371
|
+
hp->cs = cs;
|
372
|
+
hp->start.offset = p - buffer;
|
373
|
+
|
374
|
+
assert(p <= pe && "buffer overflow after parsing execute");
|
375
|
+
assert(hp->start.offset <= len && "start.offset longer than length");
|
376
|
+
}
|
377
|
+
|
378
|
+
static struct http_parser *data_get(VALUE self)
|
379
|
+
{
|
380
|
+
struct http_parser *hp;
|
381
|
+
|
382
|
+
Data_Get_Struct(self, struct http_parser, hp);
|
383
|
+
assert(hp && "failed to extract http_parser struct");
|
384
|
+
return hp;
|
385
|
+
}
|
386
|
+
|
387
|
+
static void finalize_header(struct http_parser *hp, VALUE req)
|
388
|
+
{
|
389
|
+
VALUE temp = rb_hash_aref(req, g_rack_url_scheme);
|
390
|
+
VALUE server_name = g_localhost;
|
391
|
+
VALUE server_port = g_port_80;
|
392
|
+
|
393
|
+
/* set rack.url_scheme to "https" or "http", no others are allowed by Rack */
|
394
|
+
if (NIL_P(temp)) {
|
395
|
+
temp = rb_hash_aref(req, g_http_x_forwarded_proto);
|
396
|
+
if (!NIL_P(temp) && STR_CSTR_EQ(temp, "https"))
|
397
|
+
server_port = g_port_443;
|
398
|
+
else
|
399
|
+
temp = g_http;
|
400
|
+
rb_hash_aset(req, g_rack_url_scheme, temp);
|
401
|
+
} else if (STR_CSTR_EQ(temp, "https")) {
|
402
|
+
server_port = g_port_443;
|
403
|
+
} else {
|
404
|
+
assert(server_port == g_port_80 && "server_port not set");
|
405
|
+
}
|
406
|
+
|
407
|
+
/* parse and set the SERVER_NAME and SERVER_PORT variables */
|
408
|
+
temp = rb_hash_aref(req, g_http_host);
|
409
|
+
if (!NIL_P(temp)) {
|
410
|
+
char *colon = memchr(RSTRING_PTR(temp), ':', RSTRING_LEN(temp));
|
411
|
+
if (colon) {
|
412
|
+
long port_start = colon - RSTRING_PTR(temp) + 1;
|
413
|
+
|
414
|
+
server_name = rb_str_substr(temp, 0, colon - RSTRING_PTR(temp));
|
415
|
+
if ((RSTRING_LEN(temp) - port_start) > 0)
|
416
|
+
server_port = rb_str_substr(temp, port_start, RSTRING_LEN(temp));
|
417
|
+
} else {
|
418
|
+
server_name = temp;
|
419
|
+
}
|
420
|
+
}
|
421
|
+
rb_hash_aset(req, g_server_name, server_name);
|
422
|
+
rb_hash_aset(req, g_server_port, server_port);
|
423
|
+
if (!HP_FL_TEST(hp, HASHEADER))
|
424
|
+
rb_hash_aset(req, g_server_protocol, g_http_09);
|
425
|
+
|
426
|
+
/* rack requires QUERY_STRING */
|
427
|
+
if (NIL_P(rb_hash_aref(req, g_query_string)))
|
428
|
+
rb_hash_aset(req, g_query_string, rb_str_new(NULL, 0));
|
429
|
+
}
|
430
|
+
|
431
|
+
static void hp_mark(void *ptr)
|
432
|
+
{
|
433
|
+
struct http_parser *hp = ptr;
|
434
|
+
|
435
|
+
rb_gc_mark(hp->cont);
|
436
|
+
}
|
437
|
+
|
438
|
+
static VALUE HttpParser_alloc(VALUE klass)
|
439
|
+
{
|
440
|
+
struct http_parser *hp;
|
441
|
+
return Data_Make_Struct(klass, struct http_parser, hp_mark, NULL, hp);
|
442
|
+
}
|
443
|
+
|
444
|
+
|
445
|
+
/**
|
446
|
+
* call-seq:
|
447
|
+
* parser.new => parser
|
448
|
+
*
|
449
|
+
* Creates a new parser.
|
450
|
+
*/
|
451
|
+
static VALUE HttpParser_init(VALUE self)
|
452
|
+
{
|
453
|
+
http_parser_init(data_get(self));
|
454
|
+
|
455
|
+
return self;
|
456
|
+
}
|
457
|
+
|
458
|
+
/**
|
459
|
+
* call-seq:
|
460
|
+
* parser.reset => nil
|
461
|
+
*
|
462
|
+
* Resets the parser to it's initial state so that you can reuse it
|
463
|
+
* rather than making new ones.
|
464
|
+
*/
|
465
|
+
static VALUE HttpParser_reset(VALUE self)
|
466
|
+
{
|
467
|
+
http_parser_init(data_get(self));
|
468
|
+
|
469
|
+
return Qnil;
|
470
|
+
}
|
471
|
+
|
472
|
+
static void advance_str(VALUE str, off_t nr)
|
473
|
+
{
|
474
|
+
long len = RSTRING_LEN(str);
|
475
|
+
|
476
|
+
if (len == 0)
|
477
|
+
return;
|
478
|
+
|
479
|
+
rb_str_modify(str);
|
480
|
+
|
481
|
+
assert(nr <= len && "trying to advance past end of buffer");
|
482
|
+
len -= nr;
|
483
|
+
if (len > 0) /* unlikely, len is usually 0 */
|
484
|
+
memmove(RSTRING_PTR(str), RSTRING_PTR(str) + nr, len);
|
485
|
+
rb_str_set_len(str, len);
|
486
|
+
}
|
487
|
+
|
488
|
+
/**
|
489
|
+
* call-seq:
|
490
|
+
* parser.content_length => nil or Integer
|
491
|
+
*
|
492
|
+
* Returns the number of bytes left to run through HttpParser#filter_body.
|
493
|
+
* This will initially be the value of the "Content-Length" HTTP header
|
494
|
+
* after header parsing is complete and will decrease in value as
|
495
|
+
* HttpParser#filter_body is called for each chunk. This should return
|
496
|
+
* zero for requests with no body.
|
497
|
+
*
|
498
|
+
* This will return nil on "Transfer-Encoding: chunked" requests.
|
499
|
+
*/
|
500
|
+
static VALUE HttpParser_content_length(VALUE self)
|
501
|
+
{
|
502
|
+
struct http_parser *hp = data_get(self);
|
503
|
+
|
504
|
+
return HP_FL_TEST(hp, CHUNKED) ? Qnil : OFFT2NUM(hp->len.content);
|
505
|
+
}
|
506
|
+
|
507
|
+
/**
|
508
|
+
* Document-method: trailers
|
509
|
+
* call-seq:
|
510
|
+
* parser.trailers(req, data) => req or nil
|
511
|
+
*
|
512
|
+
* This is an alias for HttpParser#headers
|
513
|
+
*/
|
514
|
+
|
515
|
+
/**
|
516
|
+
* Document-method: headers
|
517
|
+
* call-seq:
|
518
|
+
* parser.headers(req, data) => req or nil
|
519
|
+
*
|
520
|
+
* Takes a Hash and a String of data, parses the String of data filling
|
521
|
+
* in the Hash returning the Hash if parsing is finished, nil otherwise
|
522
|
+
* When returning the req Hash, it may modify data to point to where
|
523
|
+
* body processing should begin.
|
524
|
+
*
|
525
|
+
* Raises HttpParserError if there are parsing errors.
|
526
|
+
*/
|
527
|
+
static VALUE HttpParser_headers(VALUE self, VALUE req, VALUE data)
|
528
|
+
{
|
529
|
+
struct http_parser *hp = data_get(self);
|
530
|
+
|
531
|
+
rb_str_update(data);
|
532
|
+
|
533
|
+
http_parser_execute(hp, req, RSTRING_PTR(data), RSTRING_LEN(data));
|
534
|
+
VALIDATE_MAX_LENGTH(hp->start.offset, HEADER);
|
535
|
+
|
536
|
+
if (hp->cs == http_parser_first_final ||
|
537
|
+
hp->cs == http_parser_en_ChunkedBody) {
|
538
|
+
advance_str(data, hp->start.offset + 1);
|
539
|
+
hp->start.offset = 0;
|
540
|
+
|
541
|
+
return req;
|
542
|
+
}
|
543
|
+
|
544
|
+
if (hp->cs == http_parser_error)
|
545
|
+
rb_raise(eHttpParserError, "Invalid HTTP format, parsing fails.");
|
546
|
+
|
547
|
+
return Qnil;
|
548
|
+
}
|
549
|
+
|
550
|
+
static int chunked_eof(struct http_parser *hp)
|
551
|
+
{
|
552
|
+
return ((hp->cs == http_parser_first_final) || HP_FL_TEST(hp, INTRAILER));
|
553
|
+
}
|
554
|
+
|
555
|
+
/**
|
556
|
+
* call-seq:
|
557
|
+
* parser.body_eof? => true or false
|
558
|
+
*
|
559
|
+
* Detects if we're done filtering the body or not. This can be used
|
560
|
+
* to detect when to stop calling HttpParser#filter_body.
|
561
|
+
*/
|
562
|
+
static VALUE HttpParser_body_eof(VALUE self)
|
563
|
+
{
|
564
|
+
struct http_parser *hp = data_get(self);
|
565
|
+
|
566
|
+
if (HP_FL_TEST(hp, CHUNKED))
|
567
|
+
return chunked_eof(hp) ? Qtrue : Qfalse;
|
568
|
+
|
569
|
+
return hp->len.content == 0 ? Qtrue : Qfalse;
|
570
|
+
}
|
571
|
+
|
572
|
+
/**
|
573
|
+
* call-seq:
|
574
|
+
* parser.keepalive? => true or false
|
575
|
+
*
|
576
|
+
* This should be used to detect if a request can really handle
|
577
|
+
* keepalives and pipelining. Currently, the rules are:
|
578
|
+
*
|
579
|
+
* 1. MUST be a GET or HEAD request
|
580
|
+
* 2. MUST be HTTP/1.1 +or+ HTTP/1.0 with "Connection: keep-alive"
|
581
|
+
* 3. MUST NOT have "Connection: close" set
|
582
|
+
*/
|
583
|
+
static VALUE HttpParser_keepalive(VALUE self)
|
584
|
+
{
|
585
|
+
struct http_parser *hp = data_get(self);
|
586
|
+
|
587
|
+
return HP_FL_ALL(hp, KEEPALIVE) ? Qtrue : Qfalse;
|
588
|
+
}
|
589
|
+
|
590
|
+
/**
|
591
|
+
* call-seq:
|
592
|
+
* parser.headers? => true or false
|
593
|
+
*
|
594
|
+
* This should be used to detect if a request has headers (and if
|
595
|
+
* the response will have headers as well). HTTP/0.9 requests
|
596
|
+
* should return false, all subsequent HTTP versions will return true
|
597
|
+
*/
|
598
|
+
static VALUE HttpParser_has_headers(VALUE self)
|
599
|
+
{
|
600
|
+
struct http_parser *hp = data_get(self);
|
601
|
+
|
602
|
+
return HP_FL_TEST(hp, HASHEADER) ? Qtrue : Qfalse;
|
603
|
+
}
|
604
|
+
|
605
|
+
/**
|
606
|
+
* call-seq:
|
607
|
+
* parser.filter_body(buf, data) => nil/data
|
608
|
+
*
|
609
|
+
* Takes a String of +data+, will modify data if dechunking is done.
|
610
|
+
* Returns +nil+ if there is more data left to process. Returns
|
611
|
+
* +data+ if body processing is complete. When returning +data+,
|
612
|
+
* it may modify +data+ so the start of the string points to where
|
613
|
+
* the body ended so that trailer processing can begin.
|
614
|
+
*
|
615
|
+
* Raises HttpParserError if there are dechunking errors.
|
616
|
+
* Basically this is a glorified memcpy(3) that copies +data+
|
617
|
+
* into +buf+ while filtering it through the dechunker.
|
618
|
+
*/
|
619
|
+
static VALUE HttpParser_filter_body(VALUE self, VALUE buf, VALUE data)
|
620
|
+
{
|
621
|
+
struct http_parser *hp = data_get(self);
|
622
|
+
char *dptr;
|
623
|
+
long dlen;
|
624
|
+
|
625
|
+
rb_str_update(data);
|
626
|
+
dptr = RSTRING_PTR(data);
|
627
|
+
dlen = RSTRING_LEN(data);
|
628
|
+
|
629
|
+
StringValue(buf);
|
630
|
+
rb_str_resize(buf, dlen); /* we can never copy more than dlen bytes */
|
631
|
+
OBJ_TAINT(buf); /* keep weirdo $SAFE users happy */
|
632
|
+
|
633
|
+
if (HP_FL_TEST(hp, CHUNKED)) {
|
634
|
+
if (!chunked_eof(hp)) {
|
635
|
+
hp->s.dest_offset = 0;
|
636
|
+
http_parser_execute(hp, buf, dptr, dlen);
|
637
|
+
if (hp->cs == http_parser_error)
|
638
|
+
rb_raise(eHttpParserError, "Invalid HTTP format, parsing fails.");
|
639
|
+
|
640
|
+
assert(hp->s.dest_offset <= hp->start.offset &&
|
641
|
+
"destination buffer overflow");
|
642
|
+
advance_str(data, hp->start.offset);
|
643
|
+
rb_str_set_len(buf, hp->s.dest_offset);
|
644
|
+
|
645
|
+
if (RSTRING_LEN(buf) == 0 && chunked_eof(hp)) {
|
646
|
+
assert(hp->len.chunk == 0 && "chunk at EOF but more to parse");
|
647
|
+
} else {
|
648
|
+
data = Qnil;
|
649
|
+
}
|
650
|
+
}
|
651
|
+
} else {
|
652
|
+
/* no need to enter the Ragel machine for unchunked transfers */
|
653
|
+
assert(hp->len.content >= 0 && "negative Content-Length");
|
654
|
+
if (hp->len.content > 0) {
|
655
|
+
long nr = MIN(dlen, hp->len.content);
|
656
|
+
|
657
|
+
memcpy(RSTRING_PTR(buf), dptr, nr);
|
658
|
+
hp->len.content -= nr;
|
659
|
+
if (hp->len.content == 0)
|
660
|
+
hp->cs = http_parser_first_final;
|
661
|
+
advance_str(data, nr);
|
662
|
+
rb_str_set_len(buf, nr);
|
663
|
+
data = Qnil;
|
664
|
+
}
|
665
|
+
}
|
666
|
+
hp->start.offset = 0; /* for trailer parsing */
|
667
|
+
return data;
|
668
|
+
}
|
669
|
+
|
670
|
+
#define SET_GLOBAL(var,str) do { \
|
671
|
+
var = find_common_field(str, sizeof(str) - 1); \
|
672
|
+
assert(!NIL_P(var) && "missed global field"); \
|
673
|
+
} while (0)
|
674
|
+
|
675
|
+
void Init_unicorn_http(void)
|
676
|
+
{
|
677
|
+
mUnicorn = rb_define_module("Unicorn");
|
678
|
+
eHttpParserError =
|
679
|
+
rb_define_class_under(mUnicorn, "HttpParserError", rb_eIOError);
|
680
|
+
cHttpParser = rb_define_class_under(mUnicorn, "HttpParser", rb_cObject);
|
681
|
+
init_globals();
|
682
|
+
rb_define_alloc_func(cHttpParser, HttpParser_alloc);
|
683
|
+
rb_define_method(cHttpParser, "initialize", HttpParser_init,0);
|
684
|
+
rb_define_method(cHttpParser, "reset", HttpParser_reset,0);
|
685
|
+
rb_define_method(cHttpParser, "headers", HttpParser_headers, 2);
|
686
|
+
rb_define_method(cHttpParser, "filter_body", HttpParser_filter_body, 2);
|
687
|
+
rb_define_method(cHttpParser, "trailers", HttpParser_headers, 2);
|
688
|
+
rb_define_method(cHttpParser, "content_length", HttpParser_content_length, 0);
|
689
|
+
rb_define_method(cHttpParser, "body_eof?", HttpParser_body_eof, 0);
|
690
|
+
rb_define_method(cHttpParser, "keepalive?", HttpParser_keepalive, 0);
|
691
|
+
rb_define_method(cHttpParser, "headers?", HttpParser_has_headers, 0);
|
692
|
+
|
693
|
+
/*
|
694
|
+
* The maximum size a single chunk when using chunked transfer encoding.
|
695
|
+
* This is only a theoretical maximum used to detect errors in clients,
|
696
|
+
* it is highly unlikely to encounter clients that send more than
|
697
|
+
* several kilobytes at once.
|
698
|
+
*/
|
699
|
+
rb_define_const(cHttpParser, "CHUNK_MAX", OFFT2NUM(UH_OFF_T_MAX));
|
700
|
+
|
701
|
+
/*
|
702
|
+
* The maximum size of the body as specified by Content-Length.
|
703
|
+
* This is only a theoretical maximum, the actual limit is subject
|
704
|
+
* to the limits of the file system used for +Dir.tmpdir+.
|
705
|
+
*/
|
706
|
+
rb_define_const(cHttpParser, "LENGTH_MAX", OFFT2NUM(UH_OFF_T_MAX));
|
707
|
+
|
708
|
+
init_common_fields();
|
709
|
+
SET_GLOBAL(g_http_host, "HOST");
|
710
|
+
SET_GLOBAL(g_http_trailer, "TRAILER");
|
711
|
+
SET_GLOBAL(g_http_transfer_encoding, "TRANSFER_ENCODING");
|
712
|
+
SET_GLOBAL(g_content_length, "CONTENT_LENGTH");
|
713
|
+
SET_GLOBAL(g_http_connection, "CONNECTION");
|
714
|
+
}
|
715
|
+
#undef SET_GLOBAL
|