/** * @file _httpoprequest.cpp * @brief Definitions for internal class HttpOpRequest * * $LicenseInfo:firstyear=2012&license=viewerlgpl$ * Second Life Viewer Source Code * Copyright (C) 2012-2014, Linden Research, Inc. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; * version 2.1 of the License only. * * This library is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA * * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA * $/LicenseInfo$ */ #include "_httpoprequest.h" #include #include #include "httpcommon.h" #include "httphandler.h" #include "httpresponse.h" #include "bufferarray.h" #include "httpheaders.h" #include "httpoptions.h" #include "_httprequestqueue.h" #include "_httpreplyqueue.h" #include "_httpservice.h" #include "_httppolicy.h" #include "_httppolicyglobal.h" #include "_httplibcurl.h" #include "_httpinternal.h" #include "llhttpconstants.h" #include "llproxy.h" // *DEBUG: "[curl:bugs] #1420" problem and testing. // // A pipelining problem, https://sourceforge.net/p/curl/bugs/1420/, // was a source of Core_9 failures. Code related to this can be // identified and tested by: // * Looking for '[curl:bugs]' strings in source and following // instructions there. // * Set 'QAModeHttpTrace' to 2 or 3 in settings.xml and look for // 'timed out' events in the log. // * Enable the HttpRangeRequestsDisable debug setting which causes // full asset fetches. These slow the pipelines down a bit. // namespace { // Attempts to parse a 'Content-Range:' header. Caller must already // have verified that the header tag is present. The 'buffer' argument // will be processed by strtok_r calls which will modify the buffer. // // @return -1 if invalid and response should be dropped, 0 if valid an // correct, 1 if couldn't be parsed. If 0, the first, last, // and length arguments are also written. 'length' may be // 0 if the length wasn't available to the server. // int parse_content_range_header(char * buffer, unsigned int * first, unsigned int * last, unsigned int * length); // Similar for Retry-After headers. Only parses the delta form // of the header, HTTP time formats aren't interesting for client // purposes. // // @return 0 if successfully parsed and seconds time delta // returned in time argument. // int parse_retry_after_header(char * buffer, int * time); // Take data from libcurl's CURLOPT_DEBUGFUNCTION callback and // escape and format it for a tracing line in logging. Absolutely // anything including NULs can be in the data. If @scrub is true, // non-printing or non-ascii characters are replaced with spaces // otherwise a %XX form of escaping is used. void escape_libcurl_debug_data(char * buffer, size_t len, bool scrub, std::string & safe_line); // OS-neutral string comparisons of various types. int os_strcasecmp(const char * s1, const char * s2); char * os_strtok_r(char * str, const char * delim, char ** saveptr); char * os_strtrim(char * str); char * os_strltrim(char * str); void os_strlower(char * str); // Error testing and reporting for libcurl status codes void check_curl_easy_code(CURLcode code, int curl_setopt_option); static const char * const LOG_CORE("CoreHttp"); } // end anonymous namespace namespace LLCore { HttpOpRequest::HttpOpRequest() : HttpOperation(), mProcFlags(0U), mReqMethod(HOR_GET), mReqBody(NULL), mReqOffset(0), mReqLength(0), mReqHeaders(), mReqOptions(), mCurlActive(false), mCurlHandle(NULL), mCurlService(NULL), mCurlHeaders(NULL), mCurlBodyPos(0), mCurlTemp(NULL), mCurlTempLen(0), mReplyBody(NULL), mReplyOffset(0), mReplyLength(0), mReplyFullLength(0), mReplyHeaders(), mPolicyRetries(0), mPolicy503Retries(0), mPolicyRetryAt(HttpTime(0)), mPolicyRetryLimit(HTTP_RETRY_COUNT_DEFAULT), mPolicyMinRetryBackoff(HttpTime(HTTP_RETRY_BACKOFF_MIN_DEFAULT)), mPolicyMaxRetryBackoff(HttpTime(HTTP_RETRY_BACKOFF_MAX_DEFAULT)), mCallbackSSLVerify(NULL) { // *NOTE: As members are added, retry initialization/cleanup // may need to be extended in @see prepareRequest(). } HttpOpRequest::~HttpOpRequest() { if (mReqBody) { mReqBody->release(); mReqBody = NULL; } if (mCurlHandle) { // Uncertain of thread context so free using // safest method. curl_easy_cleanup(mCurlHandle); mCurlHandle = NULL; } mCurlService = NULL; if (mCurlHeaders) { curl_slist_free_all(mCurlHeaders); mCurlHeaders = NULL; } delete [] mCurlTemp; mCurlTemp = NULL; mCurlTempLen = 0; if (mReplyBody) { mReplyBody->release(); mReplyBody = NULL; } } void HttpOpRequest::stageFromRequest(HttpService * service) { HttpOpRequest::ptr_t self(boost::dynamic_pointer_cast(shared_from_this())); service->getPolicy().addOp(self); // transfers refcount } void HttpOpRequest::stageFromReady(HttpService * service) { HttpOpRequest::ptr_t self(boost::dynamic_pointer_cast(shared_from_this())); service->getTransport().addOp(self); // transfers refcount } void HttpOpRequest::stageFromActive(HttpService * service) { if (mReplyLength) { // If non-zero, we received and processed a Content-Range // header with the response. If there is received data // (and there may not be due to protocol violations, // HEAD requests, etc., see BUG-2295) Verify that what it // says is consistent with the received data. if (mReplyBody && mReplyBody->size() && mReplyLength != mReplyBody->size()) { // Not as expected, fail the request mStatus = HttpStatus(HttpStatus::LLCORE, HE_INV_CONTENT_RANGE_HDR); } } if (mCurlHeaders) { // We take these headers out of the request now as they were // allocated originally in this thread and the notifier doesn't // need them. This eliminates one source of heap moving across // threads. curl_slist_free_all(mCurlHeaders); mCurlHeaders = NULL; } // Also not needed on the other side delete [] mCurlTemp; mCurlTemp = NULL; mCurlTempLen = 0; addAsReply(); } void HttpOpRequest::visitNotifier(HttpRequest * request) { if (mUserHandler) { HttpResponse * response = new HttpResponse(); response->setStatus(mStatus); response->setBody(mReplyBody); response->setHeaders(mReplyHeaders); response->setRequestURL(mReqURL); if (mReplyOffset || mReplyLength) { // Got an explicit offset/length in response response->setRange(mReplyOffset, mReplyLength, mReplyFullLength); } response->setContentType(mReplyConType); response->setRetries(mPolicyRetries, mPolicy503Retries); HttpResponse::TransferStats::ptr_t stats = HttpResponse::TransferStats::ptr_t(new HttpResponse::TransferStats); curl_easy_getinfo(mCurlHandle, CURLINFO_SIZE_DOWNLOAD, &stats->mSizeDownload); curl_easy_getinfo(mCurlHandle, CURLINFO_TOTAL_TIME, &stats->mTotalTime); curl_easy_getinfo(mCurlHandle, CURLINFO_SPEED_DOWNLOAD, &stats->mSpeedDownload); response->setTransferStats(stats); mUserHandler->onCompleted(this->getHandle(), response); response->release(); } } // /*static*/ // HttpOpRequest::ptr_t HttpOpRequest::fromHandle(HttpHandle handle) // { // // return boost::dynamic_pointer_cast((static_cast(handle))->shared_from_this()); // } HttpStatus HttpOpRequest::cancel() { mStatus = HttpStatus(HttpStatus::LLCORE, HE_OP_CANCELED); addAsReply(); return HttpStatus(); } HttpStatus HttpOpRequest::setupGet(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t & headers) { setupCommon(policy_id, priority, url, NULL, options, headers); mReqMethod = HOR_GET; return HttpStatus(); } HttpStatus HttpOpRequest::setupGetByteRange(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, size_t offset, size_t len, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t & headers) { setupCommon(policy_id, priority, url, NULL, options, headers); mReqMethod = HOR_GET; mReqOffset = offset; mReqLength = len; if (offset || len) { mProcFlags |= PF_SCAN_RANGE_HEADER; } return HttpStatus(); } HttpStatus HttpOpRequest::setupPost(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, BufferArray * body, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t & headers) { setupCommon(policy_id, priority, url, body, options, headers); mReqMethod = HOR_POST; return HttpStatus(); } HttpStatus HttpOpRequest::setupPut(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, BufferArray * body, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t & headers) { setupCommon(policy_id, priority, url, body, options, headers); mReqMethod = HOR_PUT; return HttpStatus(); } HttpStatus HttpOpRequest::setupDelete(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t & headers) { setupCommon(policy_id, priority, url, NULL, options, headers); mReqMethod = HOR_DELETE; return HttpStatus(); } HttpStatus HttpOpRequest::setupPatch(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, BufferArray * body, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t & headers) { setupCommon(policy_id, priority, url, body, options, headers); mReqMethod = HOR_PATCH; return HttpStatus(); } HttpStatus HttpOpRequest::setupCopy(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t &headers) { setupCommon(policy_id, priority, url, NULL, options, headers); mReqMethod = HOR_COPY; return HttpStatus(); } HttpStatus HttpOpRequest::setupMove(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t &headers) { setupCommon(policy_id, priority, url, NULL, options, headers); mReqMethod = HOR_MOVE; return HttpStatus(); } void HttpOpRequest::setupCommon(HttpRequest::policy_t policy_id, HttpRequest::priority_t priority, const std::string & url, BufferArray * body, const HttpOptions::ptr_t & options, const HttpHeaders::ptr_t & headers) { mProcFlags = 0U; mReqPolicy = policy_id; mReqPriority = priority; mReqURL = url; if (body) { body->addRef(); mReqBody = body; } if (headers && ! mReqHeaders) { mReqHeaders = headers; } if (options && !mReqOptions) { mReqOptions = options; if (options->getWantHeaders()) { mProcFlags |= PF_SAVE_HEADERS; } if (options->getUseRetryAfter()) { mProcFlags |= PF_USE_RETRY_AFTER; } mPolicyRetryLimit = options->getRetries(); mPolicyRetryLimit = llclamp(mPolicyRetryLimit, HTTP_RETRY_COUNT_MIN, HTTP_RETRY_COUNT_MAX); mTracing = (std::max)(mTracing, llclamp(options->getTrace(), HTTP_TRACE_MIN, HTTP_TRACE_MAX)); mPolicyMinRetryBackoff = llclamp(options->getMinBackoff(), HttpTime(0), HTTP_RETRY_BACKOFF_MAX); mPolicyMaxRetryBackoff = llclamp(options->getMaxBackoff(), mPolicyMinRetryBackoff, HTTP_RETRY_BACKOFF_MAX); } } // Sets all libcurl options and data for a request. // // Used both for initial requests and to 'reload' for // a retry, generally with a different CURL handle. // Junk may be left around from a failed request and that // needs to be cleaned out. // // *TODO: Move this to _httplibcurl where it belongs. HttpStatus HttpOpRequest::prepareRequest(HttpService * service) { CURLcode code; // Scrub transport and result data for retried op case mCurlActive = false; mCurlHandle = NULL; mCurlService = NULL; if (mCurlHeaders) { curl_slist_free_all(mCurlHeaders); mCurlHeaders = NULL; } mCurlBodyPos = 0; if (mReplyBody) { mReplyBody->release(); mReplyBody = NULL; } mReplyOffset = 0; mReplyLength = 0; mReplyFullLength = 0; mReplyHeaders.reset(); mReplyConType.clear(); // *FIXME: better error handling later HttpStatus status; // Get global and class policy options HttpPolicyGlobal & gpolicy(service->getPolicy().getGlobalOptions()); HttpPolicyClass & cpolicy(service->getPolicy().getClassOptions(mReqPolicy)); mCurlHandle = service->getTransport().getHandle(); if (! mCurlHandle) { // We're in trouble. We'll continue but it won't go well. LL_WARNS(LOG_CORE) << "Failed to allocate libcurl easy handle. Continuing." << LL_ENDL; return HttpStatus(HttpStatus::LLCORE, HE_BAD_ALLOC); } code = curl_easy_setopt(mCurlHandle, CURLOPT_IPRESOLVE, CURL_IPRESOLVE_V4); check_curl_easy_code(code, CURLOPT_IPRESOLVE); code = curl_easy_setopt(mCurlHandle, CURLOPT_NOSIGNAL, 1); check_curl_easy_code(code, CURLOPT_NOSIGNAL); code = curl_easy_setopt(mCurlHandle, CURLOPT_NOPROGRESS, 1); check_curl_easy_code(code, CURLOPT_NOPROGRESS); code = curl_easy_setopt(mCurlHandle, CURLOPT_URL, mReqURL.c_str()); check_curl_easy_code(code, CURLOPT_URL); code = curl_easy_setopt(mCurlHandle, CURLOPT_PRIVATE, getHandle()); check_curl_easy_code(code, CURLOPT_PRIVATE); code = curl_easy_setopt(mCurlHandle, CURLOPT_ENCODING, ""); check_curl_easy_code(code, CURLOPT_ENCODING); code = curl_easy_setopt(mCurlHandle, CURLOPT_AUTOREFERER, 1); check_curl_easy_code(code, CURLOPT_AUTOREFERER); code = curl_easy_setopt(mCurlHandle, CURLOPT_MAXREDIRS, HTTP_REDIRECTS_DEFAULT); check_curl_easy_code(code, CURLOPT_MAXREDIRS); code = curl_easy_setopt(mCurlHandle, CURLOPT_WRITEFUNCTION, writeCallback); check_curl_easy_code(code, CURLOPT_WRITEFUNCTION); code = curl_easy_setopt(mCurlHandle, CURLOPT_WRITEDATA, getHandle()); check_curl_easy_code(code, CURLOPT_WRITEDATA); code = curl_easy_setopt(mCurlHandle, CURLOPT_READFUNCTION, readCallback); check_curl_easy_code(code, CURLOPT_READFUNCTION); code = curl_easy_setopt(mCurlHandle, CURLOPT_READDATA, getHandle()); check_curl_easy_code(code, CURLOPT_READDATA); code = curl_easy_setopt(mCurlHandle, CURLOPT_SEEKFUNCTION, seekCallback); check_curl_easy_code(code, CURLOPT_SEEKFUNCTION); code = curl_easy_setopt(mCurlHandle, CURLOPT_SEEKDATA, getHandle()); check_curl_easy_code(code, CURLOPT_SEEKDATA); code = curl_easy_setopt(mCurlHandle, CURLOPT_COOKIEFILE, ""); check_curl_easy_code(code, CURLOPT_COOKIEFILE); if (gpolicy.mSslCtxCallback) { code = curl_easy_setopt(mCurlHandle, CURLOPT_SSL_CTX_FUNCTION, curlSslCtxCallback); check_curl_easy_code(code, CURLOPT_SSL_CTX_FUNCTION); code = curl_easy_setopt(mCurlHandle, CURLOPT_SSL_CTX_DATA, getHandle()); check_curl_easy_code(code, CURLOPT_SSL_CTX_DATA); mCallbackSSLVerify = gpolicy.mSslCtxCallback; } long follow_redirect(1L); long sslPeerV(0L); long sslHostV(0L); long dnsCacheTimeout(-1L); long nobody(0L); if (mReqOptions) { follow_redirect = mReqOptions->getFollowRedirects() ? 1L : 0L; sslPeerV = mReqOptions->getSSLVerifyPeer() ? 1L : 0L; sslHostV = mReqOptions->getSSLVerifyHost() ? 2L : 0L; dnsCacheTimeout = mReqOptions->getDNSCacheTimeout(); nobody = mReqOptions->getHeadersOnly() ? 1L : 0L; } code = curl_easy_setopt(mCurlHandle, CURLOPT_FOLLOWLOCATION, follow_redirect); check_curl_easy_code(code, CURLOPT_FOLLOWLOCATION); code = curl_easy_setopt(mCurlHandle, CURLOPT_SSL_VERIFYPEER, sslPeerV); check_curl_easy_code(code, CURLOPT_SSL_VERIFYPEER); code = curl_easy_setopt(mCurlHandle, CURLOPT_SSL_VERIFYHOST, sslHostV); check_curl_easy_code(code, CURLOPT_SSL_VERIFYHOST); code = curl_easy_setopt(mCurlHandle, CURLOPT_NOBODY, nobody); check_curl_easy_code(code, CURLOPT_NOBODY); // The Linksys WRT54G V5 router has an issue with frequent // DNS lookups from LAN machines. If they happen too often, // like for every HTTP request, the router gets annoyed after // about 700 or so requests and starts issuing TCP RSTs to // new connections. Reuse the DNS lookups for even a few // seconds and no RSTs. code = curl_easy_setopt(mCurlHandle, CURLOPT_DNS_CACHE_TIMEOUT, dnsCacheTimeout); check_curl_easy_code(code, CURLOPT_DNS_CACHE_TIMEOUT); if (gpolicy.mUseLLProxy) { // Use the viewer-based thread-safe API which has a // fast/safe check for proxy enable. Would like to // encapsulate this someway... if (LLProxy::instanceExists()) { // Make sure proxy won't be initialized from here, // it might conflict with LLStartUp::startLLProxy() LLProxy::getInstance()->applyProxySettings(mCurlHandle); } else { LL_WARNS() << "Proxy is not initialized!" << LL_ENDL; } } else if (gpolicy.mHttpProxy.size()) { // *TODO: This is fine for now but get fuller socks5/ // authentication thing going later.... code = curl_easy_setopt(mCurlHandle, CURLOPT_PROXY, gpolicy.mHttpProxy.c_str()); check_curl_easy_code(code, CURLOPT_PROXY); code = curl_easy_setopt(mCurlHandle, CURLOPT_PROXYTYPE, CURLPROXY_HTTP); check_curl_easy_code(code, CURLOPT_PROXYTYPE); } if (gpolicy.mCAPath.size()) { code = curl_easy_setopt(mCurlHandle, CURLOPT_CAPATH, gpolicy.mCAPath.c_str()); check_curl_easy_code(code, CURLOPT_CAPATH); } if (gpolicy.mCAFile.size()) { code = curl_easy_setopt(mCurlHandle, CURLOPT_CAINFO, gpolicy.mCAFile.c_str()); check_curl_easy_code(code, CURLOPT_CAINFO); } switch (mReqMethod) { case HOR_GET: if (nobody == 0) code = curl_easy_setopt(mCurlHandle, CURLOPT_HTTPGET, 1); check_curl_easy_code(code, CURLOPT_HTTPGET); break; case HOR_POST: { code = curl_easy_setopt(mCurlHandle, CURLOPT_POST, 1); check_curl_easy_code(code, CURLOPT_POST); code = curl_easy_setopt(mCurlHandle, CURLOPT_ENCODING, ""); check_curl_easy_code(code, CURLOPT_ENCODING); long data_size(0); if (mReqBody) { data_size = mReqBody->size(); } code = curl_easy_setopt(mCurlHandle, CURLOPT_POSTFIELDS, static_cast(NULL)); check_curl_easy_code(code, CURLOPT_POSTFIELDS); code = curl_easy_setopt(mCurlHandle, CURLOPT_POSTFIELDSIZE, data_size); check_curl_easy_code(code, CURLOPT_POSTFIELDSIZE); mCurlHeaders = curl_slist_append(mCurlHeaders, "Expect:"); } break; case HOR_PATCH: code = curl_easy_setopt(mCurlHandle, CURLOPT_CUSTOMREQUEST, "PATCH"); check_curl_easy_code(code, CURLOPT_CUSTOMREQUEST); // fall through. The rest is the same as PUT case HOR_PUT: { code = curl_easy_setopt(mCurlHandle, CURLOPT_UPLOAD, 1); check_curl_easy_code(code, CURLOPT_UPLOAD); long data_size(0); if (mReqBody) { data_size = mReqBody->size(); } code = curl_easy_setopt(mCurlHandle, CURLOPT_INFILESIZE, data_size); check_curl_easy_code(code, CURLOPT_INFILESIZE); mCurlHeaders = curl_slist_append(mCurlHeaders, "Expect:"); } break; case HOR_DELETE: code = curl_easy_setopt(mCurlHandle, CURLOPT_CUSTOMREQUEST, "DELETE"); check_curl_easy_code(code, CURLOPT_CUSTOMREQUEST); break; case HOR_COPY: code = curl_easy_setopt(mCurlHandle, CURLOPT_CUSTOMREQUEST, "COPY"); check_curl_easy_code(code, CURLOPT_CUSTOMREQUEST); break; case HOR_MOVE: code = curl_easy_setopt(mCurlHandle, CURLOPT_CUSTOMREQUEST, "MOVE"); check_curl_easy_code(code, CURLOPT_CUSTOMREQUEST); break; default: LL_ERRS(LOG_CORE) << "Invalid HTTP method in request: " << int(mReqMethod) << ". Can't recover." << LL_ENDL; break; } // *TODO: Should this be 'Keep-Alive' ? mCurlHeaders = curl_slist_append(mCurlHeaders, "Connection: keep-alive"); mCurlHeaders = curl_slist_append(mCurlHeaders, "Keep-alive: 300"); // Tracing if (mTracing >= HTTP_TRACE_CURL_HEADERS) { code = curl_easy_setopt(mCurlHandle, CURLOPT_VERBOSE, 1); check_curl_easy_code(code, CURLOPT_VERBOSE); code = curl_easy_setopt(mCurlHandle, CURLOPT_DEBUGDATA, this); check_curl_easy_code(code, CURLOPT_DEBUGDATA); code = curl_easy_setopt(mCurlHandle, CURLOPT_DEBUGFUNCTION, debugCallback); check_curl_easy_code(code, CURLOPT_DEBUGFUNCTION); } // There's a CURLOPT for this now... if ((mReqOffset || mReqLength) && HOR_GET == mReqMethod) { static const char * const fmt1("Range: bytes=%lu-%lu"); static const char * const fmt2("Range: bytes=%lu-"); char range_line[64]; #if LL_WINDOWS _snprintf_s(range_line, sizeof(range_line), sizeof(range_line) - 1, (mReqLength ? fmt1 : fmt2), (unsigned long) mReqOffset, (unsigned long) (mReqOffset + mReqLength - 1)); #else if ( mReqLength ) { snprintf(range_line, sizeof(range_line), fmt1, (unsigned long) mReqOffset, (unsigned long) (mReqOffset + mReqLength - 1)); } else { snprintf(range_line, sizeof(range_line), fmt2, (unsigned long) mReqOffset); } #endif // LL_WINDOWS range_line[sizeof(range_line) - 1] = '\0'; mCurlHeaders = curl_slist_append(mCurlHeaders, range_line); } mCurlHeaders = curl_slist_append(mCurlHeaders, "Pragma:"); // Request options long timeout(HTTP_REQUEST_TIMEOUT_DEFAULT); long xfer_timeout(HTTP_REQUEST_XFER_TIMEOUT_DEFAULT); if (mReqOptions) { timeout = mReqOptions->getTimeout(); timeout = llclamp(timeout, HTTP_REQUEST_TIMEOUT_MIN, HTTP_REQUEST_TIMEOUT_MAX); xfer_timeout = mReqOptions->getTransferTimeout(); xfer_timeout = llclamp(xfer_timeout, HTTP_REQUEST_TIMEOUT_MIN, HTTP_REQUEST_TIMEOUT_MAX); } if (xfer_timeout == 0L) { xfer_timeout = timeout; } if (cpolicy.mPipelining > 1L) { // Pipelining affects both connection and transfer timeout values. // Requests that are added to a pipeling immediately have completed // their connection so the connection delay tends to be less than // the non-pipelined value. Transfers are the opposite. Transfer // timeout starts once the connection is established and completion // can be delayed due to the pipelined requests ahead. So, it's // a handwave but bump the transfer timeout up by the pipelining // depth to give some room. // // BUG-7698, BUG-7688, BUG-7694 (others). Scylla and Charybdis // situation. Operating against a CDN having service issues may // lead to requests stalling for an arbitrarily long time with only // the CURLOPT_TIMEOUT value leading to a closed connection. Sadly // for pipelining, libcurl (7.39.0 and earlier, at minimum) starts // the clock on this value as soon as a request is started down // the wire. We want a short value to recover and retry from the // CDN. We need a long value to safely deal with a succession of // piled-up pipelined requests. // // *TODO: Find a better scheme than timeouts to guarantee liveness. // Progress on the connection is what we really want, not timeouts. // But we don't have access to that and the request progress indicators // (various libcurl callbacks) have the same problem TIMEOUT does. // // xfer_timeout *= cpolicy.mPipelining; xfer_timeout *= 2L; } // *DEBUG: Enable following override for timeout handling and "[curl:bugs] #1420" tests //if (cpolicy.mPipelining) //{ // xfer_timeout = 1L; // timeout = 1L; //} code = curl_easy_setopt(mCurlHandle, CURLOPT_TIMEOUT, xfer_timeout); check_curl_easy_code(code, CURLOPT_TIMEOUT); code = curl_easy_setopt(mCurlHandle, CURLOPT_CONNECTTIMEOUT, timeout); check_curl_easy_code(code, CURLOPT_CONNECTTIMEOUT); // Request headers if (mReqHeaders) { // Caller's headers last to override mCurlHeaders = append_headers_to_slist(mReqHeaders, mCurlHeaders); } code = curl_easy_setopt(mCurlHandle, CURLOPT_HTTPHEADER, mCurlHeaders); check_curl_easy_code(code, CURLOPT_HTTPHEADER); if (mProcFlags & (PF_SCAN_RANGE_HEADER | PF_SAVE_HEADERS | PF_USE_RETRY_AFTER)) { code = curl_easy_setopt(mCurlHandle, CURLOPT_HEADERFUNCTION, headerCallback); check_curl_easy_code(code, CURLOPT_HEADERFUNCTION); code = curl_easy_setopt(mCurlHandle, CURLOPT_HEADERDATA, this); check_curl_easy_code(code, CURLOPT_HEADERDATA); } if (status) { mCurlService = service; } return status; } size_t HttpOpRequest::writeCallback(void * data, size_t size, size_t nmemb, void * userdata) { HttpOpRequest::ptr_t op(HttpOpRequest::fromHandle(userdata)); if (! op->mReplyBody) { op->mReplyBody = new BufferArray(); } const size_t req_size(size * nmemb); const size_t write_size(op->mReplyBody->append(static_cast(data), req_size)); return write_size; } size_t HttpOpRequest::readCallback(void * data, size_t size, size_t nmemb, void * userdata) { HttpOpRequest::ptr_t op(HttpOpRequest::fromHandle(userdata)); if (! op->mReqBody) { return 0; } const size_t req_size(size * nmemb); const size_t body_size(op->mReqBody->size()); if (body_size <= op->mCurlBodyPos) { if (body_size < op->mCurlBodyPos) { // Warn but continue if the read position moves beyond end-of-body // for some reason. LL_WARNS(LOG_CORE) << "Request body position beyond body size. Truncating request body." << LL_ENDL; } return 0; } const size_t do_size((std::min)(req_size, body_size - op->mCurlBodyPos)); const size_t read_size(op->mReqBody->read(op->mCurlBodyPos, static_cast(data), do_size)); op->mCurlBodyPos += read_size; return read_size; } int HttpOpRequest::seekCallback(void *userdata, curl_off_t offset, int origin) { HttpOpRequest::ptr_t op(HttpOpRequest::fromHandle(userdata)); if (!op->mReqBody) { return 0; } size_t newPos = 0; if (origin == SEEK_SET) newPos = offset; else if (origin == SEEK_END) newPos = static_cast(op->mReqBody->size()) + offset; else if (origin == SEEK_CUR) newPos = static_cast(op->mCurlBodyPos) + offset; else return 2; if (newPos >= op->mReqBody->size()) { LL_WARNS(LOG_CORE) << "Attempt to seek to position outside post body." << LL_ENDL; return 2; } op->mCurlBodyPos = (size_t)newPos; return 0; } size_t HttpOpRequest::headerCallback(void * data, size_t size, size_t nmemb, void * userdata) { static const char status_line[] = "HTTP/"; static const size_t status_line_len = sizeof(status_line) - 1; static const char con_ran_line[] = "content-range"; static const char con_retry_line[] = "retry-after"; HttpOpRequest::ptr_t op(HttpOpRequest::fromHandle(userdata)); const size_t hdr_size(size * nmemb); const char * hdr_data(static_cast(data)); // Not null terminated bool is_header(true); if (hdr_size >= status_line_len && ! strncmp(status_line, hdr_data, status_line_len)) { // One of possibly several status lines. Reset what we know and start over // taking results from the last header stanza we receive. op->mReplyOffset = 0; op->mReplyLength = 0; op->mReplyFullLength = 0; op->mReplyRetryAfter = 0; op->mStatus = HttpStatus(); if (op->mReplyHeaders) { op->mReplyHeaders->clear(); } is_header = false; } // Nothing in here wants a final CR/LF combination. Remove // it as much as possible. size_t wanted_hdr_size(hdr_size); if (wanted_hdr_size && '\n' == hdr_data[wanted_hdr_size - 1]) { if (--wanted_hdr_size && '\r' == hdr_data[wanted_hdr_size - 1]) { --wanted_hdr_size; } } // Copy and normalize header fragments for the following // stages. Would like to modify the data in-place but that // may not be allowed and we need one byte extra for NUL. // At the end of this we will have: // // If ':' present in header: // 1. name points to text to left of colon which // will be ascii lower-cased and left and right // trimmed of whitespace. // 2. value points to text to right of colon which // will be left trimmed of whitespace. // Otherwise: // 1. name points to header which will be left // trimmed of whitespace. // 2. value is NULL // Any non-NULL pointer may point to a zero-length string. // if (wanted_hdr_size >= op->mCurlTempLen) { delete [] op->mCurlTemp; op->mCurlTempLen = 2 * wanted_hdr_size + 1; op->mCurlTemp = new char [op->mCurlTempLen]; } memcpy(op->mCurlTemp, hdr_data, wanted_hdr_size); op->mCurlTemp[wanted_hdr_size] = '\0'; char * name(op->mCurlTemp); char * value(strchr(name, ':')); if (value) { *value++ = '\0'; os_strlower(name); name = os_strtrim(name); value = os_strltrim(value); } else { // Doesn't look well-formed, do minimal normalization on it name = os_strltrim(name); } // Normalized, now reject headers with empty names. if (! *name) { // No use continuing return hdr_size; } // Save header if caller wants them in the response if (is_header && op->mProcFlags & PF_SAVE_HEADERS) { // Save headers in response if (! op->mReplyHeaders) { op->mReplyHeaders = HttpHeaders::ptr_t(new HttpHeaders); } op->mReplyHeaders->append(name, value ? value : ""); } // From this point, header-specific processors are free to // modify the header value. // Detect and parse 'Content-Range' headers if (is_header && op->mProcFlags & PF_SCAN_RANGE_HEADER && value && *value && ! strcmp(name, con_ran_line)) { unsigned int first(0), last(0), length(0); int status; if (! (status = parse_content_range_header(value, &first, &last, &length))) { // Success, record the fragment position op->mReplyOffset = first; op->mReplyLength = last - first + 1; op->mReplyFullLength = length; } else if (-1 == status) { // Response is badly formed and shouldn't be accepted op->mStatus = HttpStatus(HttpStatus::LLCORE, HE_INV_CONTENT_RANGE_HDR); } else { // Ignore the unparsable. LL_INFOS_ONCE(LOG_CORE) << "Problem parsing odd Content-Range header: '" << std::string(hdr_data, wanted_hdr_size) << "'. Ignoring." << LL_ENDL; } } // Detect and parse 'Retry-After' headers if (is_header && op->mProcFlags & PF_USE_RETRY_AFTER && value && *value && ! strcmp(name, con_retry_line)) { int time(0); if (! parse_retry_after_header(value, &time)) { op->mReplyRetryAfter = time; } } return hdr_size; } CURLcode HttpOpRequest::curlSslCtxCallback(CURL *curl, void *sslctx, void *userdata) { HttpOpRequest::ptr_t op(HttpOpRequest::fromHandle(userdata)); if (op->mCallbackSSLVerify) { SSL_CTX * ctx = (SSL_CTX *)sslctx; // disable any default verification for server certs SSL_CTX_set_verify(ctx, SSL_VERIFY_NONE, NULL); // set the verification callback. SSL_CTX_set_cert_verify_callback(ctx, sslCertVerifyCallback, userdata); // the calls are void } return CURLE_OK; } int HttpOpRequest::sslCertVerifyCallback(X509_STORE_CTX *ctx, void *param) { HttpOpRequest::ptr_t op(HttpOpRequest::fromHandle(param)); if (op->mCallbackSSLVerify) { op->mStatus = op->mCallbackSSLVerify(op->mReqURL, op->mUserHandler, ctx); } return (op->mStatus) ? 1 : 0; } int HttpOpRequest::debugCallback(CURL * handle, curl_infotype info, char * buffer, size_t len, void * userdata) { HttpOpRequest::ptr_t op(HttpOpRequest::fromHandle(userdata)); std::string safe_line; std::string tag; bool logit(false); const size_t log_len((std::min)(len, size_t(256))); // Keep things reasonable in all cases switch (info) { case CURLINFO_TEXT: if (op->mTracing >= HTTP_TRACE_CURL_HEADERS) { tag = "TEXT"; escape_libcurl_debug_data(buffer, log_len, true, safe_line); logit = true; } break; case CURLINFO_HEADER_IN: if (op->mTracing >= HTTP_TRACE_CURL_HEADERS) { tag = "HEADERIN"; escape_libcurl_debug_data(buffer, log_len, true, safe_line); logit = true; } break; case CURLINFO_HEADER_OUT: if (op->mTracing >= HTTP_TRACE_CURL_HEADERS) { tag = "HEADEROUT"; escape_libcurl_debug_data(buffer, log_len, true, safe_line); // Goes out as one line unlike header_in logit = true; } break; case CURLINFO_DATA_IN: if (op->mTracing >= HTTP_TRACE_CURL_HEADERS) { tag = "DATAIN"; logit = true; if (op->mTracing >= HTTP_TRACE_CURL_BODIES) { escape_libcurl_debug_data(buffer, log_len, false, safe_line); } else { std::ostringstream out; out << len << " Bytes"; safe_line = out.str(); } } break; case CURLINFO_DATA_OUT: if (op->mTracing >= HTTP_TRACE_CURL_HEADERS) { tag = "DATAOUT"; logit = true; if (op->mTracing >= HTTP_TRACE_CURL_BODIES) { escape_libcurl_debug_data(buffer, log_len, false, safe_line); } else { std::ostringstream out; out << len << " Bytes"; safe_line = out.str(); } } break; default: logit = false; break; } if (logit) { LL_INFOS(LOG_CORE) << "TRACE, LibcurlDebug, Handle: " << op->getHandle() << ", Type: " << tag << ", Data: " << safe_line << LL_ENDL; } return 0; } } // end namespace LLCore // ======================================= // Anonymous Namespace // ======================================= namespace { int parse_content_range_header(char * buffer, unsigned int * first, unsigned int * last, unsigned int * length) { static const char * const hdr_whitespace(" \t"); char * tok_state(NULL), * tok(NULL); bool match(true); if (! (tok = os_strtok_r(buffer, hdr_whitespace, &tok_state))) match = false; else match = (0 == os_strcasecmp("bytes", tok)); if (match && ! (tok = os_strtok_r(NULL, hdr_whitespace, &tok_state))) match = false; if (match) { unsigned int lcl_first(0), lcl_last(0), lcl_len(0); #if LL_WINDOWS if (3 == sscanf_s(tok, "%u-%u/%u", &lcl_first, &lcl_last, &lcl_len)) #else if (3 == sscanf(tok, "%u-%u/%u", &lcl_first, &lcl_last, &lcl_len)) #endif // LL_WINDOWS { if (lcl_first > lcl_last || lcl_last >= lcl_len) return -1; *first = lcl_first; *last = lcl_last; *length = lcl_len; return 0; } #if LL_WINDOWS if (2 == sscanf_s(tok, "%u-%u/*", &lcl_first, &lcl_last)) #else if (2 == sscanf(tok, "%u-%u/*", &lcl_first, &lcl_last)) #endif // LL_WINDOWS { if (lcl_first > lcl_last) return -1; *first = lcl_first; *last = lcl_last; *length = 0; return 0; } } // Header is there but badly/unexpectedly formed, try to ignore it. return 1; } int parse_retry_after_header(char * buffer, int * time) { char * endptr(buffer); long lcl_time(strtol(buffer, &endptr, 10)); if (*endptr == '\0' && endptr != buffer && lcl_time > 0) { *time = lcl_time; return 0; } // Could attempt to parse HTTP time here but we're not really // interested in it. Scheduling based on wallclock time on // user hardware will lead to tears. // Header is there but badly/unexpectedly formed, try to ignore it. return 1; } void escape_libcurl_debug_data(char * buffer, size_t len, bool scrub, std::string & safe_line) { std::string out; len = (std::min)(len, size_t(200)); out.reserve(3 * len); for (int i(0); i < len; ++i) { unsigned char uc(static_cast(buffer[i])); if (uc < 32 || uc > 126) { if (scrub) { out.append(1, ' '); } else { static const char hex[] = "0123456789ABCDEF"; char convert[4]; convert[0] = '%'; convert[1] = hex[(uc >> 4) % 16]; convert[2] = hex[uc % 16]; convert[3] = '\0'; out.append(convert); } } else { out.append(1, buffer[i]); } } safe_line.swap(out); } int os_strcasecmp(const char *s1, const char *s2) { #if LL_WINDOWS return _stricmp(s1, s2); #else return strcasecmp(s1, s2); #endif // LL_WINDOWS } char * os_strtok_r(char *str, const char *delim, char ** savestate) { #if LL_WINDOWS return strtok_s(str, delim, savestate); #else return strtok_r(str, delim, savestate); #endif } void os_strlower(char * str) { for (char c(0); (c = *str); ++str) { *str = tolower(c); } } char * os_strtrim(char * lstr) { while (' ' == *lstr || '\t' == *lstr) { ++lstr; } if (*lstr) { char * rstr(lstr + strlen(lstr)); while (lstr < rstr && *--rstr) { if (' ' == *rstr || '\t' == *rstr) { *rstr = '\0'; } } llassert(lstr <= rstr); } return lstr; } char * os_strltrim(char * lstr) { while (' ' == *lstr || '\t' == *lstr) { ++lstr; } return lstr; } void check_curl_easy_code(CURLcode code, int curl_setopt_option) { if (CURLE_OK != code) { // Comment from old llcurl code which may no longer apply: // // linux appears to throw a curl error once per session for a bad initialization // at a pretty random time (when enabling cookies). LL_WARNS(LOG_CORE) << "libcurl error detected: " << curl_easy_strerror(code) << ", curl_easy_setopt option: " << curl_setopt_option << LL_ENDL; } } } // end anonymous namespace