src/HttpMsg.cc

   1
   2 /*
   3  * DEBUG: section 74    HTTP Message
   4  * AUTHOR: Alex Rousskov
   5  *
   6  * SQUID Web Proxy Cache          http://www.squid-cache.org/
   7  * ----------------------------------------------------------
   8  *
   9  *  Squid is the result of efforts by numerous individuals from
  10  *  the Internet community; see the CONTRIBUTORS file for full
  11  *  details.   Many organizations have provided support for Squid's
  12  *  development; see the SPONSORS file for full details.  Squid is
  13  *  Copyrighted (C) 2001 by the Regents of the University of
  14  *  California; see the COPYRIGHT file for full details.  Squid
  15  *  incorporates software developed and/or copyrighted by other
  16  *  sources; see the CREDITS file for full details.
  17  *
  18  *  This program is free software; you can redistribute it and/or modify
  19  *  it under the terms of the GNU General Public License as published by
  20  *  the Free Software Foundation; either version 2 of the License, or
  21  *  (at your option) any later version.
  22  *
  23  *  This program is distributed in the hope that it will be useful,
  24  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  25  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  26  *  GNU General Public License for more details.
  27  *
  28  *  You should have received a copy of the GNU General Public License
  29  *  along with this program; if not, write to the Free Software
  30  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA.
  31  *
  32  */
  33
  34 #include "squid.h"
  35 #include "Debug.h"
  36 #include "HttpHeaderTools.h"
  37 #include "HttpMsg.h"
  38 #include "MemBuf.h"
  39 #include "mime_header.h"
  40 #include "profiler/Profiler.h"
  41
  42 HttpMsg::HttpMsg(http_hdr_owner_type owner): header(owner),
  43         cache_control(NULL), hdr_sz(0), content_length(0), protocol(AnyP::PROTO_NONE),
  44         pstate(psReadyToParseStartLine), lock_count(0)
  45 {}
  46
  47 HttpMsg::~HttpMsg()
  48 {
  49     assert(lock_count == 0);
  50     assert(!body_pipe);
  51 }
  52
  53 HttpMsgParseState &operator++ (HttpMsgParseState &aState)
  54 {
  55     int tmp = (int)aState;
  56     aState = (HttpMsgParseState)(++tmp);
  57     return aState;
  58 }
  59
  60 /* find end of headers */
  61 int
  62 httpMsgIsolateHeaders(const char **parse_start, int l, const char **blk_start, const char **blk_end)
  63 {
  64     /*
  65      * parse_start points to the first line of HTTP message *headers*,
  66      * not including the request or status lines
  67      */
  68     size_t end = headersEnd(*parse_start, l);
  69     int nnl;
  70
  71     if (end) {
  72         *blk_start = *parse_start;
  73         *blk_end = *parse_start + end - 1;
  74         /*
  75          * leave blk_end pointing to the first character after the
  76          * first newline which terminates the headers
  77          */
  78         assert(**blk_end == '\n');
  79
  80         while (*(*blk_end - 1) == '\r')
  81             --(*blk_end);
  82
  83         assert(*(*blk_end - 1) == '\n');
  84
  85         *parse_start += end;
  86
  87         return 1;
  88     }
  89
  90     /*
  91      * If we didn't find the end of headers, and parse_start does
  92      * NOT point to a CR or NL character, then return failure
  93      */
  94     if (**parse_start != '\r' && **parse_start != '\n')
  95         return 0;               /* failure */
  96
  97     /*
  98      * If we didn't find the end of headers, and parse_start does point
  99      * to an empty line, then we have empty headers.  Skip all CR and
 100      * NL characters up to the first NL.  Leave parse_start pointing at
 101      * the first character after the first NL.
 102      */
 103     *blk_start = *parse_start;
 104
 105     *blk_end = *blk_start;
 106
 107     for (nnl = 0; nnl == 0; ++(*parse_start)) {
 108         if (**parse_start == '\r')
 109             (void) 0;
 110         else if (**parse_start == '\n')
 111             ++nnl;
 112         else
 113             break;
 114     }
 115
 116     return 1;
 117 }
 118
 119 /* find first CRLF */
 120 static int
 121 httpMsgIsolateStart(const char **parse_start, const char **blk_start, const char **blk_end)
 122 {
 123     int slen = strcspn(*parse_start, "\r\n");
 124
 125     if (!(*parse_start)[slen])  /* no CRLF found */
 126         return 0;
 127
 128     *blk_start = *parse_start;
 129
 130     *blk_end = *blk_start + slen;
 131
 132     while (**blk_end == '\r')   /* CR */
 133         ++(*blk_end);
 134
 135     if (**blk_end == '\n')      /* LF */
 136         ++(*blk_end);
 137
 138     *parse_start = *blk_end;
 139
 140     return 1;
 141 }
 142
 143 // negative return is the negated HTTP_ error code
 144 // zero return means need more data
 145 // positive return is the size of parsed headers
 146 bool HttpMsg::parse(MemBuf *buf, bool eof, http_status *error)
 147 {
 148     assert(error);
 149     *error = HTTP_STATUS_NONE;
 150
 151     // httpMsgParseStep() and debugging require 0-termination, unfortunately
 152     buf->terminate(); // does not affect content size
 153
 154     // find the end of headers
 155     const size_t hdr_len = headersEnd(buf->content(), buf->contentSize());
 156
 157     // sanity check the start line to see if this is in fact an HTTP message
 158     if (!sanityCheckStartLine(buf, hdr_len, error)) {
 159         // NP: sanityCheck sets *error and sends debug warnings on syntax errors.
 160         // if we have seen the connection close, this is an error too
 161         if (eof && *error==HTTP_STATUS_NONE)
 162             *error = HTTP_INVALID_HEADER;
 163
 164         return false;
 165     }
 166
 167     // TODO: move to httpReplyParseStep()
 168     if (hdr_len > Config.maxReplyHeaderSize || (hdr_len <= 0 && (size_t)buf->contentSize() > Config.maxReplyHeaderSize)) {
 169         debugs(58, DBG_IMPORTANT, "HttpMsg::parse: Too large reply header (" << hdr_len << " > " << Config.maxReplyHeaderSize);
 170         *error = HTTP_HEADER_TOO_LARGE;
 171         return false;
 172     }
 173
 174     if (hdr_len <= 0) {
 175         debugs(58, 3, "HttpMsg::parse: failed to find end of headers (eof: " << eof << ") in '" << buf->content() << "'");
 176
 177         if (eof) // iff we have seen the end, this is an error
 178             *error = HTTP_INVALID_HEADER;
 179
 180         return false;
 181     }
 182
 183     const int res = httpMsgParseStep(buf->content(), buf->contentSize(), eof);
 184
 185     if (res < 0) { // error
 186         debugs(58, 3, "HttpMsg::parse: cannot parse isolated headers in '" << buf->content() << "'");
 187         *error = HTTP_INVALID_HEADER;
 188         return false;
 189     }
 190
 191     if (res == 0) {
 192         debugs(58, 2, "HttpMsg::parse: strange, need more data near '" << buf->content() << "'");
 193         *error = HTTP_INVALID_HEADER;
 194         return false; // but this should not happen due to headersEnd() above
 195     }
 196
 197     assert(res > 0);
 198     debugs(58, 9, "HttpMsg::parse success (" << hdr_len << " bytes) near '" << buf->content() << "'");
 199
 200     if (hdr_sz != (int)hdr_len) {
 201         debugs(58, DBG_IMPORTANT, "internal HttpMsg::parse vs. headersEnd error: " <<
 202                hdr_sz << " != " << hdr_len);
 203         hdr_sz = (int)hdr_len; // because old http.cc code used hdr_len
 204     }
 205
 206     return true;
 207 }
 208
 209 /*
 210  * parseCharBuf() takes character buffer of HTTP headers (buf),
 211  * which may not be NULL-terminated, and fills in an HttpMsg
 212  * structure.  The parameter 'end' specifies the offset to
 213  * the end of the reply headers.  The caller may know where the
 214  * end is, but is unable to NULL-terminate the buffer.  This function
 215  * returns true on success.
 216  */
 217 bool
 218 HttpMsg::parseCharBuf(const char *buf, ssize_t end)
 219 {
 220     MemBuf mb;
 221     int success;
 222     /* reset current state, because we are not used in incremental fashion */
 223     reset();
 224     mb.init();
 225     mb.append(buf, end);
 226     mb.terminate();
 227     success = httpMsgParseStep(mb.buf, mb.size, 0);
 228     mb.clean();
 229     return success == 1;
 230 }
 231
 232 /*
 233  * parses a 0-terminating buffer into HttpMsg.
 234  * Returns:
 235  *      1 -- success
 236  *       0 -- need more data (partial parse)
 237  *      -1 -- parse error
 238  */
 239 int
 240 HttpMsg::httpMsgParseStep(const char *buf, int len, int atEnd)
 241 {
 242     const char *parse_start = buf;
 243     int parse_len = len;
 244     const char *blk_start, *blk_end;
 245     const char **parse_end_ptr = &blk_end;
 246     assert(parse_start);
 247     assert(pstate < psParsed);
 248
 249     *parse_end_ptr = parse_start;
 250
 251     PROF_start(HttpMsg_httpMsgParseStep);
 252
 253     if (pstate == psReadyToParseStartLine) {
 254         if (!httpMsgIsolateStart(&parse_start, &blk_start, &blk_end)) {
 255             PROF_stop(HttpMsg_httpMsgParseStep);
 256             return 0;
 257         }
 258
 259         if (!parseFirstLine(blk_start, blk_end)) {
 260             PROF_stop(HttpMsg_httpMsgParseStep);
 261             return httpMsgParseError();
 262         }
 263
 264         *parse_end_ptr = parse_start;
 265
 266         hdr_sz = *parse_end_ptr - buf;
 267         parse_len = parse_len - hdr_sz;
 268
 269         ++pstate;
 270     }
 271
 272     /*
 273      * XXX This code uses parse_start; but if we're incrementally parsing then
 274      * this code might not actually be given parse_start at the right spot (just
 275      * after headers.) Grr.
 276      */
 277     if (pstate == psReadyToParseHeaders) {
 278         if (!httpMsgIsolateHeaders(&parse_start, parse_len, &blk_start, &blk_end)) {
 279             if (atEnd) {
 280                 blk_start = parse_start, blk_end = blk_start + strlen(blk_start);
 281             } else {
 282                 PROF_stop(HttpMsg_httpMsgParseStep);
 283                 return 0;
 284             }
 285         }
 286
 287         if (!header.parse(blk_start, blk_end)) {
 288             PROF_stop(HttpMsg_httpMsgParseStep);
 289             return httpMsgParseError();
 290         }
 291
 292         hdrCacheInit();
 293
 294         *parse_end_ptr = parse_start;
 295
 296         hdr_sz = *parse_end_ptr - buf;
 297
 298         ++pstate;
 299     }
 300
 301     PROF_stop(HttpMsg_httpMsgParseStep);
 302     return 1;
 303 }
 304
 305 /* handy: resets and returns -1 */
 306 int
 307 HttpMsg::httpMsgParseError()
 308 {
 309     reset();
 310     return -1;
 311 }
 312
 313 void
 314 HttpMsg::setContentLength(int64_t clen)
 315 {
 316     header.delById(HDR_CONTENT_LENGTH); // if any
 317     header.putInt64(HDR_CONTENT_LENGTH, clen);
 318     content_length = clen;
 319 }
 320
 321 bool
 322 HttpMsg::persistent() const
 323 {
 324     if (http_ver > HttpVersion(1, 0)) {
 325         /*
 326          * for modern versions of HTTP: persistent unless there is
 327          * a "Connection: close" header.
 328          */
 329         return !httpHeaderHasConnDir(&header, "close");
 330     } else {
 331         /* for old versions of HTTP: persistent if has "keep-alive" */
 332         return httpHeaderHasConnDir(&header, "keep-alive");
 333     }
 334 }
 335
 336 void HttpMsg::packInto(Packer *p, bool full_uri) const
 337 {
 338     packFirstLineInto(p, full_uri);
 339     header.packInto(p);
 340     packerAppend(p, "\r\n", 2);
 341 }
 342
 343 void HttpMsg::hdrCacheInit()
 344 {
 345     content_length = header.getInt64(HDR_CONTENT_LENGTH);
 346     assert(NULL == cache_control);
 347     cache_control = header.getCc();
 348 }
 349
 350 /*
 351  * useful for debugging
 352  */
 353 void HttpMsg::firstLineBuf(MemBuf& mb)
 354 {
 355     Packer p;
 356     packerToMemInit(&p, &mb);
 357     packFirstLineInto(&p, true);
 358     packerClean(&p);
 359 }
 360
 361 // use HTTPMSGLOCK() instead of calling this directly
 362 HttpMsg *
 363 HttpMsg::_lock()
 364 {
 365     ++lock_count;
 366     return this;
 367 }
 368
 369 // use HTTPMSGUNLOCK() instead of calling this directly
 370 void
 371 HttpMsg::_unlock()
 372 {
 373     assert(lock_count > 0);
 374     --lock_count;
 375
 376     if (0 == lock_count)
 377         delete this;
 378 }