xmltooling/util/CurlURLInputStream.cpp

   1 /*
   2  * Licensed to the Apache Software Foundation (ASF) under one or more
   3  * contributor license agreements.  See the NOTICE file distributed with
   4  * this work for additional information regarding copyright ownership.
   5  * The ASF licenses this file to You under the Apache License, Version 2.0
   6  * (the "License"); you may not use this file except in compliance with
   7  * the License.  You may obtain a copy of the License at
   8  *
   9  *      http://www.apache.org/licenses/LICENSE-2.0
  10  *
  11  * Unless required by applicable law or agreed to in writing, software
  12  * distributed under the License is distributed on an "AS IS" BASIS,
  13  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  14  * See the License for the specific language governing permissions and
  15  * limitations under the License.
  16  */
  17
  18 /*
  19  * $Id$
  20  */
  21
  22 #include "internal.h"
  23
  24 #include <stdio.h>
  25 #include <stdlib.h>
  26 #include <string.h>
  27 #include <errno.h>
  28 #ifdef HAVE_UNISTD_H
  29 # include <unistd.h>
  30 #endif
  31 //#include <sys/types.h>
  32 //#include <sys/time.h>
  33
  34 #include <xercesc/util/XercesDefs.hpp>
  35 #include <xercesc/util/XMLNetAccessor.hpp>
  36 #include <xercesc/util/XMLString.hpp>
  37 #include <xercesc/util/XMLExceptMsgs.hpp>
  38 #include <xercesc/util/Janitor.hpp>
  39 #include <xercesc/util/XMLUniDefs.hpp>
  40 #include <xercesc/util/TransService.hpp>
  41 #include <xercesc/util/TranscodingException.hpp>
  42 #include <xercesc/util/PlatformUtils.hpp>
  43
  44 #include <xmltooling/util/CurlURLInputStream.hpp>
  45
  46 using namespace xmltooling;
  47
  48
  49 CurlURLInputStream::CurlURLInputStream(const XMLURL& urlSource, const XMLNetHTTPInfo* httpInfo/*=0*/)
  50       : fMulti(0)
  51       , fEasy(0)
  52       , fMemoryManager(urlSource.getMemoryManager())
  53       , fURLSource(urlSource)
  54       , fURL(0)
  55       , fTotalBytesRead(0)
  56       , fWritePtr(0)
  57       , fBytesRead(0)
  58       , fBytesToRead(0)
  59       , fDataAvailable(false)
  60       , fBufferHeadPtr(fBuffer)
  61       , fBufferTailPtr(fBuffer)
  62       , m_log(logging::Category::getInstance(XMLTOOLING_LOGCAT".libcurl.NetAccessor"))
  63 {
  64         // Allocate the curl multi handle
  65         fMulti = curl_multi_init();
  66
  67         // Allocate the curl easy handle
  68         fEasy = curl_easy_init();
  69
  70         // Get the text of the URL we're going to use
  71         fURL.reset(XMLString::transcode(fURLSource.getURLText(), fMemoryManager), fMemoryManager);
  72
  73         m_log.debug("libcurl trying to fetch %s", fURL.get());
  74
  75         // Set URL option
  76         curl_easy_setopt(fEasy, CURLOPT_URL, fURL.get());
  77         curl_easy_setopt(fEasy, CURLOPT_WRITEDATA, this);                                               // Pass this pointer to write function
  78         curl_easy_setopt(fEasy, CURLOPT_WRITEFUNCTION, staticWriteCallback);    // Our static write function
  79     curl_easy_setopt(fEasy, CURLOPT_CONNECTTIMEOUT, 30);
  80     curl_easy_setopt(fEasy, CURLOPT_TIMEOUT, 60);
  81     curl_easy_setopt(fEasy, CURLOPT_SSL_VERIFYHOST, 0);
  82     curl_easy_setopt(fEasy, CURLOPT_SSL_VERIFYPEER, 0);
  83     curl_easy_setopt(fEasy, CURLOPT_NOPROGRESS, 1);
  84     curl_easy_setopt(fEasy, CURLOPT_NOSIGNAL, 1);
  85     curl_easy_setopt(fEasy, CURLOPT_FAILONERROR, 1);
  86
  87         // Add easy handle to the multi stack
  88         curl_multi_add_handle(fMulti, fEasy);
  89 }
  90
  91
  92 CurlURLInputStream::~CurlURLInputStream()
  93 {
  94         // Remove the easy handle from the multi stack
  95         curl_multi_remove_handle(fMulti, fEasy);
  96
  97         // Cleanup the easy handle
  98         curl_easy_cleanup(fEasy);
  99
 100         // Cleanup the multi handle
 101         curl_multi_cleanup(fMulti);
 102 }
 103
 104
 105 size_t
 106 CurlURLInputStream::staticWriteCallback(char *buffer,
 107                                       size_t size,
 108                                       size_t nitems,
 109                                       void *outstream)
 110 {
 111         return ((CurlURLInputStream*)outstream)->writeCallback(buffer, size, nitems);
 112 }
 113
 114
 115
 116 size_t
 117 CurlURLInputStream::writeCallback(char *buffer,
 118                                       size_t size,
 119                                       size_t nitems)
 120 {
 121         size_t cnt = size * nitems;
 122         size_t totalConsumed = 0;
 123
 124         // Consume as many bytes as possible immediately into the buffer
 125         size_t consume = (cnt > fBytesToRead) ? fBytesToRead : cnt;
 126         memcpy(fWritePtr, buffer, consume);
 127         fWritePtr               += consume;
 128         fBytesRead              += consume;
 129         fTotalBytesRead += consume;
 130         fBytesToRead    -= consume;
 131
 132         //m_log.debug("write callback consuming %d bytes", consume);
 133
 134         // If bytes remain, rebuffer as many as possible into our holding buffer
 135         buffer                  += consume;
 136         totalConsumed   += consume;
 137         cnt                             -= consume;
 138         if (cnt > 0)
 139         {
 140                 size_t bufAvail = sizeof(fBuffer) - (fBufferHeadPtr - fBuffer);
 141                 consume = (cnt > bufAvail) ? bufAvail : cnt;
 142                 memcpy(fBufferHeadPtr, buffer, consume);
 143                 fBufferHeadPtr  += consume;
 144                 buffer                  += consume;
 145                 totalConsumed   += consume;
 146                 //m_log.debug("write callback rebuffering %d bytes", consume);
 147         }
 148
 149         // Return the total amount we've consumed. If we don't consume all the bytes
 150         // then an error will be generated. Since our buffer size is equal to the
 151         // maximum size that curl will write, this should never happen unless there
 152         // is a logic error somewhere here.
 153         return totalConsumed;
 154 }
 155
 156
 157 unsigned int
 158 CurlURLInputStream::readBytes(XMLByte* const          toFill
 159                                      , const unsigned int maxToRead)
 160 {
 161         fBytesRead = 0;
 162         fBytesToRead = maxToRead;
 163         fWritePtr = toFill;
 164
 165         for (bool tryAgain = true; fBytesToRead > 0 && (tryAgain || fBytesRead == 0); )
 166         {
 167                 // First, any buffered data we have available
 168                 size_t bufCnt = fBufferHeadPtr - fBufferTailPtr;
 169                 bufCnt = (bufCnt > fBytesToRead) ? fBytesToRead : bufCnt;
 170                 if (bufCnt > 0)
 171                 {
 172                         memcpy(fWritePtr, fBufferTailPtr, bufCnt);
 173                         fWritePtr               += bufCnt;
 174                         fBytesRead              += bufCnt;
 175                         fTotalBytesRead += bufCnt;
 176                         fBytesToRead    -= bufCnt;
 177
 178                         fBufferTailPtr  += bufCnt;
 179                         if (fBufferTailPtr == fBufferHeadPtr)
 180                                 fBufferHeadPtr = fBufferTailPtr = fBuffer;
 181
 182                         //m_log.debug("consuming %d buffered bytes", bufCnt);
 183
 184                         tryAgain = true;
 185                         continue;
 186                 }
 187
 188                 // Ask the curl to do some work
 189                 int runningHandles = 0;
 190                 CURLMcode curlResult = curl_multi_perform(fMulti, &runningHandles);
 191         //m_log.debug("curl_multi_perform returned %d", curlResult);
 192                 tryAgain = (curlResult == CURLM_CALL_MULTI_PERFORM);
 193
 194                 // Process messages from curl
 195                 int msgsInQueue = 0;
 196                 for (CURLMsg* msg = NULL; (msg = curl_multi_info_read(fMulti, &msgsInQueue)) != NULL; )
 197                 {
 198                         m_log.debug("msg %d, %d from curl", msg->msg, msg->data.result);
 199
 200                         if (msg->msg != CURLMSG_DONE)
 201                                 continue;
 202
 203                         switch (msg->data.result)
 204                         {
 205                         case CURLE_OK:
 206                                 // We completed successfully. runningHandles should have dropped to zero, so we'll bail out below...
 207                                 break;
 208
 209                         case CURLE_UNSUPPORTED_PROTOCOL:
 210                 ThrowXMLwithMemMgr(MalformedURLException, XMLExcepts::URL_UnsupportedProto, fMemoryManager);
 211                 break;
 212
 213             case CURLE_COULDNT_RESOLVE_HOST:
 214             case CURLE_COULDNT_RESOLVE_PROXY:
 215                 ThrowXMLwithMemMgr1(NetAccessorException,  XMLExcepts::NetAcc_TargetResolution, fURLSource.getHost(), fMemoryManager);
 216                 break;
 217
 218             case CURLE_COULDNT_CONNECT:
 219                 ThrowXMLwithMemMgr1(NetAccessorException, XMLExcepts::NetAcc_ConnSocket, fURLSource.getURLText(), fMemoryManager);
 220
 221             case CURLE_RECV_ERROR:
 222                 ThrowXMLwithMemMgr1(NetAccessorException, XMLExcepts::NetAcc_ReadSocket, fURLSource.getURLText(), fMemoryManager);
 223                 break;
 224
 225             default:
 226                 m_log.error("curl NetAccessor encountered error from libcurl (%d)", msg->data.result);
 227                 ThrowXMLwithMemMgr1(NetAccessorException, XMLExcepts::NetAcc_InternalError, fURLSource.getURLText(), fMemoryManager);
 228                                 break;
 229                         }
 230                 }
 231
 232                 // If nothing is running any longer, bail out
 233         if (runningHandles == 0) {
 234             //m_log.debug("libcurl indicated no running handles");
 235                         break;
 236         }
 237
 238                 // If there is no further data to read, and we haven't
 239                 // read any yet on this invocation, call select to wait for data
 240                 if (!tryAgain && fBytesRead == 0)
 241                 {
 242                         fd_set readSet;
 243                         fd_set writeSet;
 244                         fd_set exceptSet;
 245                         int fdcnt=0;
 246
 247                         // Ask curl for the file descriptors to wait on
 248             FD_ZERO(&readSet);
 249             FD_ZERO(&writeSet);
 250             FD_ZERO(&exceptSet);
 251                         (void) curl_multi_fdset(fMulti, &readSet, &writeSet, &exceptSet, &fdcnt);
 252
 253                         // Wait on the file descriptors
 254                         timeval tv;
 255                         tv.tv_sec  = 2;
 256                         tv.tv_usec = 0;
 257                         (void) select(fdcnt+1, &readSet, &writeSet, &exceptSet, &tv);
 258                 }
 259         }
 260
 261     //m_log.debug("returning with %d bytes to parser", fBytesRead);
 262         return fBytesRead;
 263 }
 264