emulationstation-de/es-core/src/HttpReq.cpp

424 lines
14 KiB
C++

// SPDX-License-Identifier: MIT
//
// ES-DE Frontend
// HttpReq.cpp
//
// HTTP requests using libcurl.
// Used by the scraper and application updater.
//
#include "HttpReq.h"
#include "ApplicationVersion.h"
#include "Log.h"
#include "Settings.h"
#include "resources/ResourceManager.h"
#include "utils/FileSystemUtil.h"
#include <algorithm>
#include <assert.h>
std::string HttpReq::urlEncode(const std::string& s)
{
const std::string unreserved {
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_.~"};
std::string escaped {""};
for (size_t i {0}; i < s.length(); ++i) {
if (unreserved.find_first_of(s[i]) != std::string::npos) {
escaped.push_back(s[i]);
}
else {
escaped.append("%");
char buf[3];
snprintf(buf, 3, "%.2X", static_cast<unsigned char>(s[i]));
escaped.append(buf);
}
}
return escaped;
}
HttpReq::HttpReq(const std::string& url, bool scraperRequest)
: mStatus {REQ_IN_PROGRESS}
, mHandle {nullptr}
, mTotalBytes {0}
, mDownloadedBytes {0}
, mScraperRequest {scraperRequest}
{
// The multi-handle is cleaned up via an explicit call to cleanupCurlMulti() from any object
// that uses HttpReq. For example from GuiScraperSearch after scraping has been completed.
if (!sMultiHandle)
sMultiHandle = curl_multi_init();
mHandle = curl_easy_init();
if (mHandle == nullptr) {
mStatus = REQ_IO_ERROR;
onError("curl_easy_init failed");
return;
}
if (!mPollThread) {
sStopPoll = false;
mPollThread = std::make_unique<std::thread>(&HttpReq::pollCurl, this);
}
#if defined(USE_BUNDLED_CERTIFICATES)
// Use the bundled curl TLS/SSL certificates (which come from the Mozilla project).
// This is used on Windows and also on Android as there is no way for curl to access
// the system certificates on this OS.
curl_easy_setopt(mHandle, CURLOPT_CAINFO,
ResourceManager::getInstance()
.getResourcePath(":/certificates/curl-ca-bundle.crt")
.c_str());
#endif
// Set the URL.
CURLcode err {curl_easy_setopt(mHandle, CURLOPT_URL, url.c_str())};
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
if (!mScraperRequest) {
// Set User-Agent.
std::string userAgent {"ES-DE Frontend/"};
userAgent.append(PROGRAM_VERSION_STRING).append(" (");
#if defined(__ANDROID__)
userAgent.append("Android");
#elif defined(_WIN64)
userAgent.append("Windows");
#elif defined(__APPLE__)
userAgent.append("macOS");
#elif defined(__linux__)
userAgent.append("Linux");
#elif defined(__unix__)
userAgent.append("Unix");
#else
userAgent.append("Unknown");
#endif
userAgent.append(")");
CURLcode err {curl_easy_setopt(mHandle, CURLOPT_USERAGENT, userAgent.c_str())};
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
}
long connectionTimeout;
if (mScraperRequest) {
connectionTimeout =
static_cast<long>(Settings::getInstance()->getInt("ScraperConnectionTimeout"));
if (connectionTimeout < 0 || connectionTimeout > 300)
connectionTimeout = static_cast<long>(
Settings::getInstance()->getDefaultInt("ScraperConnectionTimeout"));
}
else {
connectionTimeout = 30;
}
// Set connection timeout (default is 30 seconds).
err = curl_easy_setopt(mHandle, CURLOPT_CONNECTTIMEOUT, connectionTimeout);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
long transferTimeout;
if (mScraperRequest) {
transferTimeout =
static_cast<long>(Settings::getInstance()->getInt("ScraperTransferTimeout"));
if (transferTimeout < 0 || transferTimeout > 300)
transferTimeout =
static_cast<long>(Settings::getInstance()->getDefaultInt("ScraperTransferTimeout"));
}
else {
transferTimeout = 0;
}
// Set transfer timeout (default is 120 seconds for the scraper and infinite otherwise).
err = curl_easy_setopt(mHandle, CURLOPT_TIMEOUT, transferTimeout);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Set curl to handle redirects.
err = curl_easy_setopt(mHandle, CURLOPT_FOLLOWLOCATION, 1L);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Set curl max redirects.
err = curl_easy_setopt(mHandle, CURLOPT_MAXREDIRS, 2L);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Set curl restrict redirect protocols.
#if defined(__APPLE__) || LIBCURL_VERSION_MAJOR < 7 || \
(LIBCURL_VERSION_MAJOR == 7 && LIBCURL_VERSION_MINOR < 85)
err = curl_easy_setopt(mHandle, CURLOPT_REDIR_PROTOCOLS, CURLPROTO_HTTP | CURLPROTO_HTTPS);
#else
err = curl_easy_setopt(mHandle, CURLOPT_REDIR_PROTOCOLS_STR, "http,https");
#endif
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Tell curl how to write the data.
err = curl_easy_setopt(mHandle, CURLOPT_WRITEFUNCTION, &HttpReq::writeContent);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Pass curl a pointer to this HttpReq so we know where to write the data to in our
// write function.
err = curl_easy_setopt(mHandle, CURLOPT_WRITEDATA, this);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Enable the curl progress meter.
err = curl_easy_setopt(mHandle, CURLOPT_NOPROGRESS, mScraperRequest ? 1 : 0);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Pass curl a pointer to HttpReq to provide access to the counter variables.
err = curl_easy_setopt(mHandle, CURLOPT_XFERINFODATA, this);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Progress meter callback.
if (!mScraperRequest) {
err = curl_easy_setopt(mHandle, CURLOPT_XFERINFOFUNCTION, HttpReq::transferProgress);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
}
// Fail on HTTP status codes >= 400.
err = curl_easy_setopt(mHandle, CURLOPT_FAILONERROR, 1L);
if (err != CURLE_OK) {
mStatus = REQ_IO_ERROR;
onError(curl_easy_strerror(err));
return;
}
// Add the handle to the multi. This is done in pollCurl(), running in a separate thread.
std::unique_lock<std::mutex> handleLock {sHandleMutex};
sAddHandleQueue.push(mHandle);
handleLock.unlock();
curl_multi_wakeup(sMultiHandle);
std::unique_lock<std::mutex> requestLock {sRequestMutex};
sRequests[mHandle] = this;
requestLock.unlock();
}
HttpReq::~HttpReq()
{
if (mHandle) {
std::unique_lock<std::mutex> requestLock {sRequestMutex};
sRequests.erase(mHandle);
requestLock.unlock();
std::unique_lock<std::mutex> handleLock {sHandleMutex};
sRemoveHandleQueue.push(mHandle);
handleLock.unlock();
curl_multi_wakeup(sMultiHandle);
}
}
std::string HttpReq::getContent() const
{
assert(mStatus == REQ_SUCCESS);
return mContent.str();
}
int HttpReq::transferProgress(
void* clientp, curl_off_t dltotal, curl_off_t dlnow, curl_off_t ultotal, curl_off_t ulnow)
{
if (dltotal == 0 && dlnow == 0)
return CURLE_OK;
// We need all the check logic below to make sure we're not attempting to write into
// a request that has just been removed by the main thread.
bool validEntry {false};
std::unique_lock<std::mutex> requestLock {sRequestMutex};
if (std::find_if(sRequests.cbegin(), sRequests.cend(), [&clientp](auto&& entry) {
return entry.second == clientp;
}) != sRequests.cend())
validEntry = true;
if (validEntry) {
// Note that it's not guaranteed that the server will actually provide the total size.
if (dltotal > 0)
static_cast<HttpReq*>(clientp)->mTotalBytes = static_cast<long>(dltotal);
if (dlnow > 0)
static_cast<HttpReq*>(clientp)->mDownloadedBytes = static_cast<long>(dlnow);
}
requestLock.unlock();
return CURLE_OK;
}
size_t HttpReq::writeContent(void* buff, size_t size, size_t nmemb, void* req_ptr)
{
// We need all the check logic below to make sure we're not attempting to write into
// a request that has just been removed by the main thread.
bool validEntry {false};
std::unique_lock<std::mutex> requestLock {sRequestMutex};
if (std::find_if(sRequests.cbegin(), sRequests.cend(), [&req_ptr](auto&& entry) {
return entry.second == req_ptr;
}) != sRequests.cend())
validEntry = true;
if (validEntry) {
// size = size of an element, nmemb = number of elements.
std::stringstream& ss {static_cast<HttpReq*>(req_ptr)->mContent};
ss.write(static_cast<char*>(buff), size * nmemb);
}
requestLock.unlock();
// Return value is number of elements successfully read.
return nmemb;
}
void HttpReq::pollCurl()
{
int numfds {0};
do {
if (!sStopPoll)
curl_multi_poll(sMultiHandle, nullptr, 0, 2000, &numfds);
// Check if any easy handles should be added or removed.
std::unique_lock<std::mutex> handleLock {sHandleMutex};
if (sAddHandleQueue.size() > 0) {
// Add the handle to our multi.
CURLMcode merr {curl_multi_add_handle(sMultiHandle, sAddHandleQueue.front())};
std::unique_lock<std::mutex> requestLock {sRequestMutex};
HttpReq* req {sRequests[sAddHandleQueue.front()]};
if (merr != CURLM_OK) {
if (req != nullptr) {
req->mStatus = REQ_IO_ERROR;
req->onError(curl_multi_strerror(merr));
LOG(LogError) << "onError(): " << curl_multi_strerror(merr);
}
}
else {
if (req != nullptr)
req->mStatus = REQ_IN_PROGRESS;
}
sAddHandleQueue.pop();
requestLock.unlock();
}
if (sRemoveHandleQueue.size() > 0) {
// Remove the handle from our multi.
CURLMcode merr {curl_multi_remove_handle(sMultiHandle, sRemoveHandleQueue.front())};
if (merr != CURLM_OK) {
LOG(LogError) << "Error removing curl easy handle from curl multi: "
<< curl_multi_strerror(merr);
}
curl_easy_cleanup(sRemoveHandleQueue.front());
sRemoveHandleQueue.pop();
}
handleLock.unlock();
if (sMultiHandle != nullptr && !sStopPoll) {
int handleCount {0};
std::unique_lock<std::mutex> handleLock {sHandleMutex};
CURLMcode merr {curl_multi_perform(sMultiHandle, &handleCount)};
handleLock.unlock();
if (merr != CURLM_OK && merr != CURLM_CALL_MULTI_PERFORM) {
LOG(LogError) << "Error reading data from multi: " << curl_multi_strerror(merr);
}
int msgsLeft;
CURLMsg* msg;
while (!sStopPoll && (msg = curl_multi_info_read(sMultiHandle, &msgsLeft)) != nullptr) {
if (msg->msg == CURLMSG_DONE) {
std::unique_lock<std::mutex> requestLock {sRequestMutex};
HttpReq* req {sRequests[msg->easy_handle]};
if (req == nullptr) {
LOG(LogError) << "Cannot find easy handle!";
requestLock.unlock();
continue;
}
if (msg->data.result == CURLE_OK) {
req->mStatus = REQ_SUCCESS;
}
else if (msg->data.result == CURLE_PEER_FAILED_VERIFICATION) {
req->mStatus = REQ_FAILED_VERIFICATION;
req->onError(curl_easy_strerror(msg->data.result));
}
else if (msg->data.result == CURLE_HTTP_RETURNED_ERROR) {
long responseCode;
curl_easy_getinfo(msg->easy_handle, CURLINFO_RESPONSE_CODE, &responseCode);
if (responseCode == 430 &&
Settings::getInstance()->getString("Scraper") == "screenscraper") {
req->mContent << "You have exceeded your daily scrape quota";
req->mStatus = REQ_SUCCESS;
}
else if (responseCode == 404 && req->mScraperRequest &&
Settings::getInstance()->getBool("ScraperIgnoreHTTP404Errors")) {
req->mStatus = REQ_RESOURCE_NOT_FOUND;
}
else {
req->mStatus = REQ_BAD_STATUS_CODE;
req->onError("Server returned HTTP error code " +
std::to_string(responseCode));
}
}
else {
req->mStatus = REQ_IO_ERROR;
req->onError(curl_easy_strerror(msg->data.result));
}
requestLock.unlock();
}
}
}
} while (!sStopPoll || !sAddHandleQueue.empty() || !sRemoveHandleQueue.empty());
}