2 * Copyright (C) 1996-2015 The Squid Software Foundation and contributors
4 * Squid software is distributed under GPLv2+ license and includes
5 * contributions from numerous individuals and organizations.
6 * Please see the COPYING and CONTRIBUTORS files for details.
9 /* DEBUG: section 74 HTTP Message */
13 #include "HttpHeaderTools.h"
16 #include "mime_header.h"
17 #include "profiler/Profiler.h"
18 #include "SquidConfig.h"
20 HttpMsg::HttpMsg(http_hdr_owner_type owner
):
21 http_ver(Http::ProtocolVersion()),
26 pstate(psReadyToParseStartLine
)
34 HttpMsgParseState
&operator++ (HttpMsgParseState
&aState
)
36 int tmp
= (int)aState
;
37 aState
= (HttpMsgParseState
)(++tmp
);
41 /* find end of headers */
43 httpMsgIsolateHeaders(const char **parse_start
, int l
, const char **blk_start
, const char **blk_end
)
46 * parse_start points to the first line of HTTP message *headers*,
47 * not including the request or status lines
49 size_t end
= headersEnd(*parse_start
, l
);
53 *blk_start
= *parse_start
;
54 *blk_end
= *parse_start
+ end
- 1;
56 * leave blk_end pointing to the first character after the
57 * first newline which terminates the headers
59 assert(**blk_end
== '\n');
61 while (*(*blk_end
- 1) == '\r')
64 assert(*(*blk_end
- 1) == '\n');
72 * If we didn't find the end of headers, and parse_start does
73 * NOT point to a CR or NL character, then return failure
75 if (**parse_start
!= '\r' && **parse_start
!= '\n')
76 return 0; /* failure */
79 * If we didn't find the end of headers, and parse_start does point
80 * to an empty line, then we have empty headers. Skip all CR and
81 * NL characters up to the first NL. Leave parse_start pointing at
82 * the first character after the first NL.
84 *blk_start
= *parse_start
;
86 *blk_end
= *blk_start
;
88 for (nnl
= 0; nnl
== 0; ++(*parse_start
)) {
89 if (**parse_start
== '\r')
91 else if (**parse_start
== '\n')
100 /* find first CRLF */
102 httpMsgIsolateStart(const char **parse_start
, const char **blk_start
, const char **blk_end
)
104 int slen
= strcspn(*parse_start
, "\r\n");
106 if (!(*parse_start
)[slen
]) /* no CRLF found */
109 *blk_start
= *parse_start
;
111 *blk_end
= *blk_start
+ slen
;
113 while (**blk_end
== '\r') /* CR */
116 if (**blk_end
== '\n') /* LF */
119 *parse_start
= *blk_end
;
124 // negative return is the negated Http::StatusCode error code
125 // zero return means need more data
126 // positive return is the size of parsed headers
128 HttpMsg::parse(MemBuf
*buf
, bool eof
, Http::StatusCode
*error
)
131 *error
= Http::scNone
;
133 // httpMsgParseStep() and debugging require 0-termination, unfortunately
134 buf
->terminate(); // does not affect content size
136 // find the end of headers
137 const size_t hdr_len
= headersEnd(buf
->content(), buf
->contentSize());
139 // sanity check the start line to see if this is in fact an HTTP message
140 if (!sanityCheckStartLine(buf
, hdr_len
, error
)) {
141 // NP: sanityCheck sets *error and sends debug warnings on syntax errors.
142 // if we have seen the connection close, this is an error too
143 if (eof
&& *error
== Http::scNone
)
144 *error
= Http::scInvalidHeader
;
149 // TODO: move to httpReplyParseStep()
150 if (hdr_len
> Config
.maxReplyHeaderSize
|| (hdr_len
<= 0 && (size_t)buf
->contentSize() > Config
.maxReplyHeaderSize
)) {
151 debugs(58, DBG_IMPORTANT
, "HttpMsg::parse: Too large reply header (" << hdr_len
<< " > " << Config
.maxReplyHeaderSize
);
152 *error
= Http::scHeaderTooLarge
;
157 debugs(58, 3, "HttpMsg::parse: failed to find end of headers (eof: " << eof
<< ") in '" << buf
->content() << "'");
159 if (eof
) // iff we have seen the end, this is an error
160 *error
= Http::scInvalidHeader
;
165 const int res
= httpMsgParseStep(buf
->content(), buf
->contentSize(), eof
);
167 if (res
< 0) { // error
168 debugs(58, 3, "HttpMsg::parse: cannot parse isolated headers in '" << buf
->content() << "'");
169 *error
= Http::scInvalidHeader
;
174 debugs(58, 2, "HttpMsg::parse: strange, need more data near '" << buf
->content() << "'");
175 *error
= Http::scInvalidHeader
;
176 return false; // but this should not happen due to headersEnd() above
180 debugs(58, 9, "HttpMsg::parse success (" << hdr_len
<< " bytes) near '" << buf
->content() << "'");
182 if (hdr_sz
!= (int)hdr_len
) {
183 debugs(58, DBG_IMPORTANT
, "internal HttpMsg::parse vs. headersEnd error: " <<
184 hdr_sz
<< " != " << hdr_len
);
185 hdr_sz
= (int)hdr_len
; // because old http.cc code used hdr_len
192 * parseCharBuf() takes character buffer of HTTP headers (buf),
193 * which may not be NULL-terminated, and fills in an HttpMsg
194 * structure. The parameter 'end' specifies the offset to
195 * the end of the reply headers. The caller may know where the
196 * end is, but is unable to NULL-terminate the buffer. This function
197 * returns true on success.
200 HttpMsg::parseCharBuf(const char *buf
, ssize_t end
)
204 /* reset current state, because we are not used in incremental fashion */
209 success
= httpMsgParseStep(mb
.buf
, mb
.size
, 0);
215 * parses a 0-terminating buffer into HttpMsg.
218 * 0 -- need more data (partial parse)
222 HttpMsg::httpMsgParseStep(const char *buf
, int len
, int atEnd
)
224 const char *parse_start
= buf
;
226 const char *blk_start
, *blk_end
;
227 const char **parse_end_ptr
= &blk_end
;
229 assert(pstate
< psParsed
);
231 *parse_end_ptr
= parse_start
;
233 PROF_start(HttpMsg_httpMsgParseStep
);
235 if (pstate
== psReadyToParseStartLine
) {
236 if (!httpMsgIsolateStart(&parse_start
, &blk_start
, &blk_end
)) {
237 PROF_stop(HttpMsg_httpMsgParseStep
);
241 if (!parseFirstLine(blk_start
, blk_end
)) {
242 PROF_stop(HttpMsg_httpMsgParseStep
);
243 return httpMsgParseError();
246 *parse_end_ptr
= parse_start
;
248 hdr_sz
= *parse_end_ptr
- buf
;
249 parse_len
= parse_len
- hdr_sz
;
255 * XXX This code uses parse_start; but if we're incrementally parsing then
256 * this code might not actually be given parse_start at the right spot (just
257 * after headers.) Grr.
259 if (pstate
== psReadyToParseHeaders
) {
260 if (!httpMsgIsolateHeaders(&parse_start
, parse_len
, &blk_start
, &blk_end
)) {
262 blk_start
= parse_start
;
263 blk_end
= blk_start
+ strlen(blk_start
);
265 PROF_stop(HttpMsg_httpMsgParseStep
);
270 if (!header
.parse(blk_start
, blk_end
-blk_start
)) {
271 PROF_stop(HttpMsg_httpMsgParseStep
);
272 return httpMsgParseError();
277 *parse_end_ptr
= parse_start
;
279 hdr_sz
= *parse_end_ptr
- buf
;
284 PROF_stop(HttpMsg_httpMsgParseStep
);
288 /* handy: resets and returns -1 */
290 HttpMsg::httpMsgParseError()
297 HttpMsg::setContentLength(int64_t clen
)
299 header
.delById(HDR_CONTENT_LENGTH
); // if any
300 header
.putInt64(HDR_CONTENT_LENGTH
, clen
);
301 content_length
= clen
;
305 HttpMsg::persistent() const
307 if (http_ver
> Http::ProtocolVersion(1,0)) {
309 * for modern versions of HTTP: persistent unless there is
310 * a "Connection: close" header.
312 return !httpHeaderHasConnDir(&header
, "close");
314 /* for old versions of HTTP: persistent if has "keep-alive" */
315 return httpHeaderHasConnDir(&header
, "keep-alive");
319 void HttpMsg::packInto(Packer
*p
, bool full_uri
) const
321 packFirstLineInto(p
, full_uri
);
323 p
->append("\r\n", 2);
326 void HttpMsg::hdrCacheInit()
328 content_length
= header
.getInt64(HDR_CONTENT_LENGTH
);
329 assert(NULL
== cache_control
);
330 cache_control
= header
.getCc();
334 * useful for debugging
336 void HttpMsg::firstLineBuf(MemBuf
& mb
)
339 packerToMemInit(&p
, &mb
);
340 packFirstLineInto(&p
, true);