boinc/client/http_curl.C

1058 lines
35 KiB
C

// Berkeley Open Infrastructure for Network Computing
// http://boinc.berkeley.edu
// Copyright (C) 2005 University of California
//
// This 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;
// either version 2.1 of the License, or (at your option) any later version.
//
// This software 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.
//
// To view the GNU Lesser General Public License visit
// http://www.gnu.org/copyleft/lesser.html
// or write to the Free Software Foundation, Inc.,
// 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
#include "cpp.h"
#ifdef _WIN32
#include "boinc_win.h"
#else
#include "config.h"
#include <cstring>
#include <sstream>
#include <algorithm>
#include <sys/stat.h>
#include <cerrno>
#include <unistd.h>
#ifdef HAVE_SYS_SOCKET_H
#include <sys/socket.h>
#endif
#endif
#include "error_numbers.h"
#include "filesys.h"
#include "client_msgs.h"
#include "log_flags.h"
#include "util.h"
#include "network.h"
#include "client_msgs.h"
#include "base64.h"
#include "http_curl.h"
#include "client_state.h"
#define CONNECTED_STATE_NOT_CONNECTED 0
#define CONNECTED_STATE_CONNECTED 1
#define CONNECTED_STATE_UNKNOWN 2
#define HTTP_BLOCKSIZE 16384
using std::min;
using std::vector;
static CURLM* g_curlMulti = NULL;
static char g_user_agent_string[256] = {""};
//static const char g_content_type[] = {"Content-Type: application/octet-stream"};
// CMC Note: old BOINC used the above, but libcurl seems to like the following:
static const char g_content_type[] = {"Content-Type: application/x-www-form-urlencoded"};
/** as an example of usage of http_op -- the scheduler typically requires
just the following:
http_op
http_op_state flag
http_op_retval
set_proxy (g_state.proxy_info)
init_get()
init_post()
http_ops
-> insert
remove
*/
// Breaks a HTTP URL down into its server, port and file components
// format of url:
// [http[s]://]host.dom.dom[:port][/dir/file]
//
void parse_url(const char* url, char* host, int &port, char* file) {
char* p;
char buf[256];
bool bSSL = false;
//const short csiLen = bSSL ? 8 : 7; // get right size of http/s string comparator
// strip off http:// if present
//
//if (strncmp(url, (bSSL ? "https://" : "http://"), csiLen) == 0) {
if (strncmp(url, "http://", 7) == 0) {
safe_strcpy(buf, url+7);
} else {
// wait, may be https://
if (strncmp(url, "https://", 8) == 0) {
safe_strcpy(buf, url+8);
bSSL = true; // retain the fact that this was a secure http url
}
else { // no http:// or https:// prepended on url
safe_strcpy(buf, url);
}
}
// parse and strip off file part if present
//
p = strchr(buf, '/');
if (p) {
strcpy(file, p+1);
*p = 0;
} else {
strcpy(file, "");
}
// parse and strip off port if present
//
p = strchr(buf,':');
if (p) {
port = atol(p+1);
*p = 0;
} else {
// CMC note: if they didn't pass in a port #,
// but the url starts with https://, assume they
// want a secure port (HTTPS, port 443)
port = (bSSL ? 443 : 80);
}
// what remains is the host
//
strcpy(host, buf);
}
void get_user_agent_string() {
sprintf(g_user_agent_string, "BOINC client (%s %d.%d.%d)",
HOSTTYPE, BOINC_MAJOR_VERSION, BOINC_MINOR_VERSION, BOINC_RELEASE
);
}
HTTP_OP::HTTP_OP() {
strcpy(m_url, "");
content_length = 0;
file_offset = 0;
strcpy(request_header, "");
http_op_state = HTTP_STATE_IDLE;
http_op_type = HTTP_OP_NONE;
http_op_retval = 0;
trace_id = 0;
pcurlList = NULL; // these have to be NULL, just in constructor
curlEasy = NULL;
pcurlFormStart = NULL;
pcurlFormEnd = NULL;
pByte = NULL;
lSeek = 0;
auth_flag = false;
auth_type = 0;
reset();
}
HTTP_OP::~HTTP_OP() {
close_socket();
close_file();
}
// Initialize HTTP GET operation
//
int HTTP_OP::init_get(
const char* url, const char* out, bool del_old_file, double off
) {
//char proxy_buf[256];
if (del_old_file) {
unlink(out);
}
req1 = NULL; // not using req1, but init_post2 uses it
file_offset = off;
HTTP_OP::init();
// usually have an outfile on a get
if (off != 0) {
bytes_xferred = off;
start_bytes_xferred = off;
}
http_op_type = HTTP_OP_GET;
http_op_state = HTTP_STATE_CONNECTING;
if (log_flags.http_debug) {
msg_printf(0, MSG_INFO, "[http_debug] HTTP_OP::init_get(): %s", url);
}
return HTTP_OP::libcurl_exec(url, NULL, out, off, false);
}
// Initialize HTTP POST operation
//
int HTTP_OP::init_post(
const char* url, const char* in, const char* out
) {
int retval;
double size;
req1 = NULL; // not using req1, but init_post2 uses it
if (in) {
// we should pretty much always have an in file for _post, optional in _post2
strcpy(infile, in);
retval = file_size(infile, size);
if (retval) return retval; // this will return 0 or ERR_NOT_FOUND
content_length = (int)size;
}
HTTP_OP::init();
http_op_type = HTTP_OP_POST;
http_op_state = HTTP_STATE_CONNECTING;
if (log_flags.http_debug) {
msg_printf(0, MSG_INFO, "[http_debug] HTTP_OP::init_post(): %s", url);
}
return HTTP_OP::libcurl_exec(url, in, out, 0.0, true); // note that no offset for this, for resumable uploads use post2!
}
// the following will do an HTTP GET or POST using libcurl
//
int HTTP_OP::libcurl_exec(
const char* url, const char* in, const char* out, double offset, bool bPost
) {
CURLMcode curlMErr;
CURLcode curlErr;
char strTmp[128];
safe_strcpy(m_url, url);
if (g_user_agent_string[0] == 0x00) {
get_user_agent_string();
}
if (in) {
strcpy(infile, in);
}
if (out) {
bTempOutfile = false;
strcpy(outfile, out);
} else {
//CMC -- I always want an outfile for the server response, delete when op done
bTempOutfile = true;
strcpy(outfile, "");
#if defined(_WIN32) && !defined(__CYGWIN32__)
char* ptrName;
ptrName = _tempnam("./", "blc");
if (ptrName) {
strcpy(outfile, ptrName);
free(ptrName);
}
#elif defined( __EMX__)
strcpy(outfile, "blcXXXXXX"); // a template for the mktemp
// mktemp will not open the file
mktemp(outfile);
#else // use mkstemp on Mac & Linux due to security issues
strcpy(outfile, "blcXXXXXX"); // a template for the mkstemp
close(mkstemp(outfile));
#endif
}
// setup libcurl handle
// CMC -- init the curlEasy handle and setup options
// the polling will do the actual start of the HTTP/S transaction
curlEasy = curl_easy_init(); // get a curl_easy handle to use
if (!curlEasy) {
msg_printf(0, MSG_ERROR, "Couldn't create curlEasy handle");
return ERR_HTTP_ERROR; // returns 0 (CURLM_OK) on successful handle creation
}
// the following seems to be a no-op
//curlErr = curl_easy_setopt(curlEasy, CURLOPT_ERRORBUFFER, error_msg);
// OK, we have a handle, now open an asynchronous libcurl connection
// set the URL to use
curlErr = curl_easy_setopt(curlEasy, CURLOPT_URL, m_url);
/*
CURLOPT_SSL_VERIFYHOST
Pass a long as parameter.
This option determines whether curl verifies that the server claims to be who you want it to be.
When negotiating an SSL connection, the server sends a certificate indicating its identity.
When CURLOPT_SSL_VERIFYHOST is 2, that certificate must indicate that the server is the server to which you meant to connect, or the connection fails.
Curl considers the server the intended one when the Common Name field or a Subject Alternate Name field in the certificate matches the host name in the URL to which you told Curl to connect.
When the value is 1, the certificate must contain a Common Name field, but it doesn't matter what name it says. (This is not ordinarily a useful setting).
When the value is 0, the connection succeeds regardless of the names in the certificate.
The default, since 7.10, is 2.
The checking this option controls is of the identity that the server claims. The server could be lying. To control lying, see CURLOPT_SSL_VERIFYPEER.
*/
curlErr = curl_easy_setopt(curlEasy, CURLOPT_SSL_VERIFYHOST, 2L);
// the following sets "tough" certificate checking
// (i.e. whether self-signed is OK)
// if zero below, will accept self-signed certificates
// (cert not 3rd party trusted)
// if non-zero below, you need a valid 3rd party CA (i.e. Verisign, Thawte)
//
curlErr = curl_easy_setopt(curlEasy, CURLOPT_SSL_VERIFYPEER, 1L);
// if the above is nonzero, you need the following:
curlErr = curl_easy_setopt(curlEasy, CURLOPT_CAINFO, "ca-bundle.crt");
// set the user agent as this boinc client & version
curlErr = curl_easy_setopt(curlEasy, CURLOPT_USERAGENT, g_user_agent_string);
// bypass any signal handlers that curl may want to install
curlErr = curl_easy_setopt(curlEasy, CURLOPT_NOSIGNAL, 1L);
// bypass progress meter
curlErr = curl_easy_setopt(curlEasy, CURLOPT_NOPROGRESS, 1L);
// setup timeouts
curlErr = curl_easy_setopt(curlEasy, CURLOPT_TIMEOUT, 0L);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_LOW_SPEED_LIMIT, 10L);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_LOW_SPEED_TIME, 300L);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_CONNECTTIMEOUT, 120L);
// force curl to use HTTP/1.1
curlErr = curl_easy_setopt(curlEasy, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_1_1);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_MAXREDIRS, 50L);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_AUTOREFERER, 1L);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_FOLLOWLOCATION, 1L);
// if we tell Curl to accept any encoding (e.g. deflate)
// it seems to accept them all, which screws up projects that
// use gzip at the application level.
// So, detect this and don't accept any encoding in that case
//
if (!out || !ends_with(std::string(out), std::string(".gz"))) {
curlErr = curl_easy_setopt(curlEasy, CURLOPT_ENCODING, "");
}
// setup any proxy they may need
setupProxyCurl();
// set the content type in the header (defined at the top -- app/octect-stream)
pcurlList = curl_slist_append(pcurlList, g_content_type);
// set the file offset for resumable downloads
if (!bPost && offset>0.0f) {
file_offset = offset;
sprintf(strTmp, "Range: bytes=%.0f-", offset);
pcurlList = curl_slist_append(pcurlList, strTmp);
}
// we'll need to setup an output file for the reply
if (outfile && strlen(outfile)>0) {
// if offset>0 don't truncate!
char szType[3] = "wb";
if (file_offset>0.0) szType[0] = 'a';
fileOut = boinc_fopen(outfile, szType);
if (!fileOut) {
msg_printf(NULL, MSG_ERROR,
"Can't create HTTP response output file %s", outfile
);
http_op_retval = ERR_FOPEN;
http_op_state = HTTP_STATE_DONE;
return ERR_FOPEN;
}
// CMC Note: we can make the libcurl_write "fancier" in the future,
// for now it just fwrite's to the file request, which is sufficient
curlErr = curl_easy_setopt(curlEasy, CURLOPT_WRITEFUNCTION, libcurl_write);
// note that in my lib_write I'm sending in a pointer to this instance of HTTP_OP
curlErr = curl_easy_setopt(curlEasy, CURLOPT_WRITEDATA, this);
}
if (bPost) { // POST
want_upload = true;
want_download = false;
if (infile && strlen(infile)>0) {
fileIn = boinc_fopen(infile, "rb");
if (!fileIn) {
msg_printf(NULL, MSG_ERROR, "No HTTP input file %s", infile);
http_op_retval = ERR_FOPEN;
http_op_state = HTTP_STATE_DONE;
return ERR_FOPEN;
}
}
if (pcurlList) { // send custom headers if required
curlErr = curl_easy_setopt(curlEasy, CURLOPT_HTTPHEADER, pcurlList);
}
// set the data file info to read for the PUT/POST
// note the use of this curl typedef for large filesizes
/* HTTP PUT method
curl_off_t fs = (curl_off_t) content_length;
curlErr = curl_easy_setopt(curlEasy, CURLOPT_POSTFIELDS, NULL);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_INFILESIZE, content_length);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_READDATA, fileIn);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_INFILESIZE_LARGE, fs);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PUT, 1L);
*/
// HTTP POST method
// set the multipart form for the file -- boinc just has the one section (file)
/* CMC -- if we ever want to do POST as multipart forms someday
// (many seem to prefer it that way, i.e. libcurl)
pcurlFormStart = pcurlFormEnd = NULL;
curl_formadd(&pcurlFormStart, &pcurlFormEnd,
CURLFORM_FILECONTENT, infile,
CURLFORM_CONTENTSLENGTH, content_length,
CURLFORM_CONTENTTYPE, g_content_type,
CURLFORM_END);
curl_formadd(&post, &last,
CURLFORM_COPYNAME, "logotype-image",
CURLFORM_FILECONTENT, "curl.png", CURLFORM_END);
*/
//curlErr = curl_easy_setopt(curlEasy, CURLOPT_HTTPPOST, pcurlFormStart);
curl_off_t fs = (curl_off_t) content_length;
pByte = NULL;
lSeek = 0; // initialize the vars we're going to use for byte transfers
// CMC Note: we can make the libcurl_read "fancier" in the future,
// for now it just fwrite's to the file request, which is sufficient
curlErr = curl_easy_setopt(curlEasy, CURLOPT_POSTFIELDS, NULL);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_POSTFIELDSIZE_LARGE, fs);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_READFUNCTION, libcurl_read);
// note that in my lib_write I'm sending in a pointer to this instance of HTTP_OP
curlErr = curl_easy_setopt(curlEasy, CURLOPT_READDATA, this);
// callback function to rewind input file
curlErr = curl_easy_setopt(curlEasy, CURLOPT_IOCTLFUNCTION, libcurl_ioctl);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_IOCTLDATA, this);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_POST, 1L);
} else { // GET
want_upload = false;
want_download = true;
// now write the header, pcurlList gets freed in net_xfer_curl
if (pcurlList) { // send custom headers if required
curlErr = curl_easy_setopt(curlEasy, CURLOPT_HTTPHEADER, pcurlList);
}
// setup the GET!
curlErr = curl_easy_setopt(curlEasy, CURLOPT_HTTPGET, 1L);
}
// turn on debug info if tracing enabled
if (log_flags.http_xfer_debug) {
static int trace_count = 0;
curlErr = curl_easy_setopt(curlEasy, CURLOPT_DEBUGFUNCTION, libcurl_debugfunction);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_DEBUGDATA, this );
curlErr = curl_easy_setopt(curlEasy, CURLOPT_VERBOSE, 1L);
trace_id = trace_count++;
}
// last but not least, add this to the curl_multi
curlMErr = curl_multi_add_handle(g_curlMulti, curlEasy);
if (curlMErr != CURLM_OK && curlMErr != CURLM_CALL_MULTI_PERFORM) { // bad error, couldn't attach easy curl handle
msg_printf(0, MSG_ERROR, "Couldn't add curlEasy handle to curlMulti");
return ERR_HTTP_ERROR; // returns 0 (CURLM_OK) on successful handle creation
}
return 0;
}
// Initialize HTTP POST operation
//
int HTTP_OP::init_post2(
const char* url, char* r1, const char* in, double offset
) {
int retval;
double size;
//char proxy_buf[256];
init();
req1 = r1;
if (in) {
safe_strcpy(infile, in);
file_offset = offset;
retval = file_size(infile, size);
if (retval) {
printf("HTTP::init_post2: couldn't get file size\n");
return retval; // this will be 0 or ERR_NOT_FOUND
}
content_length = (int)size - (int)offset;
}
content_length += (int)strlen(req1);
http_op_type = HTTP_OP_POST2;
http_op_state = HTTP_STATE_CONNECTING;
return HTTP_OP::libcurl_exec(url, in, NULL, offset, true);
}
// Returns true if the HTTP operation is complete
//
bool HTTP_OP::http_op_done() {
return (http_op_state == HTTP_STATE_DONE);
}
HTTP_OP_SET::HTTP_OP_SET() {
bytes_up = 0;
bytes_down = 0;
}
// Adds an HTTP_OP to the set
//
int HTTP_OP_SET::insert(HTTP_OP* ho) {
http_ops.push_back(ho);
return 0;
}
// Remove an HTTP_OP from the set
//
int HTTP_OP_SET::remove(HTTP_OP* p) {
vector<HTTP_OP*>::iterator iter;
iter = http_ops.begin();
while (iter != http_ops.end()) {
if (*iter == p) {
iter = http_ops.erase(iter);
return 0;
}
iter++;
}
msg_printf(NULL, MSG_ERROR, "HTTP operation not found");
return ERR_NOT_FOUND;
}
int HTTP_OP_SET::nops() {
return (int)http_ops.size();
}
int HTTP_OP::set_proxy(PROXY_INFO *new_pi) {
pi.use_http_proxy = new_pi->use_http_proxy;
strcpy(pi.http_user_name, new_pi->http_user_name);
strcpy(pi.http_user_passwd, new_pi->http_user_passwd);
strcpy(pi.http_server_name, new_pi->http_server_name);
pi.http_server_port = new_pi->http_server_port;
pi.use_http_auth = new_pi->use_http_auth;
pi.use_socks_proxy = new_pi->use_socks_proxy;
strcpy(pi.socks5_user_name, new_pi->socks5_user_name);
strcpy(pi.socks5_user_passwd, new_pi->socks5_user_passwd);
strcpy(pi.socks_server_name, new_pi->socks_server_name);
pi.socks_server_port = new_pi->socks_server_port;
pi.socks_version = new_pi->socks_version;
return 0;
}
size_t libcurl_write(void *ptr, size_t size, size_t nmemb, HTTP_OP* phop) {
// take the stream param as a FILE* and write to disk
//CMC TODO: maybe assert stRead == size*nmemb, add exception handling on phop members
size_t stWrite = fwrite(ptr, size, nmemb, (FILE*) phop->fileOut);
#if 1
if (log_flags.http_xfer_debug) {
msg_printf(NULL, MSG_INFO, "[http_xfer_debug] HTTP: wrote %d bytes", (int)stWrite);
}
#endif
phop->bytes_xferred += (double)(stWrite);
phop->update_speed(); // this should update the transfer speed
return stWrite;
}
size_t libcurl_read( void *ptr, size_t size, size_t nmemb, HTTP_OP* phop) {
// OK here's the deal -- phop points to the calling object,
// which has already pre-opened the file. we'll want to
// use pByte as a pointer for fseek calls into the file, and
// write out size*nmemb # of bytes to ptr
// take the stream param as a FILE* and write to disk
//if (pByte) delete [] pByte;
//pByte = new unsigned char[content_length];
//memset(pByte, 0x00, content_length); // may as will initialize it!
// note that fileIn was opened earlier, go to lSeek from the top and read from there
size_t stSend = size * nmemb;
int stRead = 0;
// if (phop->http_op_type == HTTP_OP_POST2) {
if (phop->req1 && ! phop->bSentHeader) { // need to send headers first, then data file
// uck -- the way 'post2' is done, you have to read the
// header bytes, and then cram on the file upload bytes
// so requests from 0 to strlen(req1)-1 are from memory,
// and from strlen(req1) to content_length are from the file
// just send the headers from htp->req1 if needed
if (phop->lSeek < (long) strlen(phop->req1)) {
// need to read header, either just starting to read (i.e.
// this is the first time in this function for this phop)
// or the last read didn't ask for the entire header
stRead = (int)strlen(phop->req1) - phop->lSeek; // how much of header left to read
// only memcpy if request isn't out of bounds
if (stRead < 0) {
stRead = 0;
} else {
memcpy(ptr, (void*)(phop->req1 + phop->lSeek), stRead);
}
phop->lSeek += (long) stRead; // increment lSeek to new position
phop->bytes_xferred += (double)(stRead);
// see if we're done with headers
phop->bSentHeader = (bool)(phop->lSeek >= (long) strlen(phop->req1));
// reset lSeek if done to make it easier for file operations
if (phop->bSentHeader) phop->lSeek = 0;
return stRead; //
}
}
// now for file to read in (if any), also don't bother if this request
// was just enough for the header (which was taken care of above)
if (phop->fileIn) {
// note we'll have to fudge lSeek a little if there was
// also a header, just use a temp var
//size_t stOld = stRead; // we'll want to save the ptr location of last stRead
// keep a separate pointer to "bump ahead" the pointer for the file data
// as ptr may have been used above for the header info
//unsigned char *ptr2;
// get the file seek offset, both from the offset requested (added)
// as well as the size of the header above discounted
// - ((phop->req1 && stRead>0) ? stRead :
// (phop->req1 ? strlen(phop->req1) : 0L))
long lFileSeek = phop->lSeek + (long) phop->file_offset;
fseek(phop->fileIn, lFileSeek, SEEK_SET);
if (!feof(phop->fileIn)) { // CMC TODO: better error checking for size*nmemb
// i.e. that we don't go overbounds of the file etc, we can check against
// content_length (which includes the strlen(req1) also)
// note the use of stOld to get to the right position in case the header was read in above
//ptr2 = (unsigned char*)ptr +(int)stOld;
stRead = (int)fread(ptr, 1, stSend, phop->fileIn);
}
phop->lSeek += (long) stRead; // increment lSeek to new position
phop->bytes_xferred += (double)(stRead);
}
phop->update_speed(); // this should update the transfer speed
return stRead;
}
curlioerr libcurl_ioctl(CURL*, curliocmd cmd, HTTP_OP* phop) {
// reset input stream to beginning - resends header
// and restarts data back to starting point
switch(cmd) {
case CURLIOCMD_RESTARTREAD:
phop->lSeek = 0;
phop->bytes_xferred = phop->file_offset;
phop->bSentHeader = false;
break;
default: // should never get here
return CURLIOE_UNKNOWNCMD;
}
return CURLIOE_OK;
}
int libcurl_debugfunction(
CURL*, curl_infotype type,
unsigned char *data, size_t size, HTTP_OP* phop
) {
const char *text;
char hdr[100];
char buf[1024];
size_t mysize;
switch (type) {
case CURLINFO_TEXT:
if (log_flags.http_debug) {
msg_printf(0, MSG_INFO,
"[http_debug] [ID#%i] info: %s\n", phop->trace_id, data
);
}
return 0;
case CURLINFO_HEADER_OUT:
text = "Sent header to server:";
break;
case CURLINFO_HEADER_IN:
text = "Received header from server:";
break;
default: /* in case a new one is introduced to shock us */
return 0;
}
sprintf( hdr,"[ID#%i] %s", phop->trace_id, text);
mysize = min(size, sizeof(buf)-1);
strncpy(buf, (char *)data, mysize);
buf[mysize]='\0';
if (log_flags.http_debug) {
msg_printf(0, MSG_INFO,
"[http_debug] %s %s\n", hdr, buf
);
}
return 0;
}
void HTTP_OP::setupProxyCurl() {
// CMC: use the libcurl proxy routines with this object's proxy information struct
/* PROXY_INFO pi useful members:
pi.http_server_name
pi.http_server_port
pi.http_user_name
pi.http_user_passwd
pi.socks5_user_name
pi.socks5_user_passwd
pi.socks_server_name
pi.socks_server_port
pi.socks_version
pi.use_http_auth
pi.use_http_proxy
pi.use_socks_proxy
Curl self-explanatory setopt params for proxies:
CURLOPT_HTTPPROXYTUNNEL
CURLOPT_PROXYTYPE (pass in CURLPROXY_HTTP or CURLPROXY_SOCKS5)
CURLOPT_PROXYPORT -- a long port #
CURLOPT_PROXY - pass in char* of the proxy url
CURLOPT_PROXYUSERPWD -- a char* in the format username:password
CURLOPT_HTTPAUTH -- pass in one of CURLAUTH_BASIC, CURLAUTH_DIGEST,
CURLAUTH_GSSNEGOTIATE, CURLAUTH_NTLM, CURLAUTH_ANY, CURLAUTH_ANYSAFE
CURLOPT_PROXYAUTH -- "or" | the above bitmasks -- only basic, digest, ntlm work
*/
CURLcode curlErr;
// CMC Note: the string szCurlProxyUserPwd must remain in memory
// outside of this method (libcurl relies on it later when it makes
// the proxy connection), so it has been placed as a member data for HTTP_OP
memset(szCurlProxyUserPwd,0x00,128);
if (pi.use_http_proxy) {
// setup a basic http proxy
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYTYPE, CURLPROXY_HTTP);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYPORT, (long) pi.http_server_port);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXY, (char*) pi.http_server_name);
if (pi.use_http_auth) {
/* testing!
fprintf(stdout, "Using httpauth for proxy: %s:%d %s:%s\n",
pi.http_server_name, pi.http_server_port,
pi.http_user_name, pi.http_user_passwd);
*/
auth_flag = true;
if (auth_type) {
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYAUTH, auth_type);
} else {
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYAUTH, CURLAUTH_ANY & ~CURLAUTH_NTLM);
}
sprintf(szCurlProxyUserPwd, "%s:%s", pi.http_user_name, pi.http_user_passwd);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYUSERPWD, szCurlProxyUserPwd);
}
} else {
if (pi.use_socks_proxy) {
//pi.socks_version -- picks between socks5 & socks4 -- but libcurl only socks5!
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYTYPE, CURLPROXY_SOCKS5);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYPORT, (long) pi.socks_server_port);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXY, (char*) pi.socks_server_name);
// libcurl uses blocking sockets with socks proxy, so limit timeout.
// - imlemented with local patch to libcurl
curlErr = curl_easy_setopt(curlEasy, CURLOPT_CONNECTTIMEOUT, 20L);
if (
strlen(pi.socks5_user_passwd)>0 || strlen(pi.socks5_user_name)>0
) {
auth_flag = false;
sprintf(szCurlProxyUserPwd, "%s:%s", pi.socks5_user_name, pi.socks5_user_passwd);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYUSERPWD, szCurlProxyUserPwd);
curlErr = curl_easy_setopt(curlEasy, CURLOPT_PROXYAUTH, CURLAUTH_ANY & ~CURLAUTH_NTLM);
}
}
}
}
// the file descriptor sets need to be global so libcurl has access always
//
fd_set read_fds, write_fds, error_fds;
// call these once at the start of the program and once at the end
//
int curl_init() {
curl_global_init(CURL_GLOBAL_ALL);
g_curlMulti = curl_multi_init();
return (int)(g_curlMulti == NULL);
}
int curl_cleanup() {
if (g_curlMulti) {
curl_multi_cleanup(g_curlMulti);
}
return 0;
}
void HTTP_OP::reset() {
req1 = NULL;
strcpy(infile, "");
strcpy(outfile, "");
strcpy(error_msg, "");
CurlResult = CURLE_OK;
bTempOutfile = true;
want_download = false;
want_upload = false;
fileIn = NULL;
fileOut = NULL;
connect_error = 0;
bytes_xferred = 0;
xfer_speed = 0;
bSentHeader = false;
close_socket();
}
void HTTP_OP::close_socket() {
// this cleans up the curlEasy, and "spoofs" the old close_socket
//
if (pcurlList) {
curl_slist_free_all(pcurlList);
pcurlList = NULL;
}
if (curlEasy && pcurlFormStart) {
curl_formfree(pcurlFormStart);
curl_formfree(pcurlFormEnd);
pcurlFormStart = pcurlFormEnd = NULL;
}
if (curlEasy && g_curlMulti) { // release this handle
curl_multi_remove_handle(g_curlMulti, curlEasy);
curl_easy_cleanup(curlEasy);
curlEasy = NULL;
}
}
void HTTP_OP::close_file() {
if (fileIn) {
fclose(fileIn);
fileIn = NULL;
}
if (fileOut) {
fclose(fileOut);
fileOut = NULL;
}
if (pByte) { //free any read memory used
delete [] pByte;
pByte = NULL;
}
}
void HTTP_OP::init() {
reset();
start_time = gstate.now;
start_bytes_xferred = 0;
}
void HTTP_OP_SET::get_fdset(FDSET_GROUP& fg) {
CURLMcode curlMErr;
curlMErr = curl_multi_fdset(
g_curlMulti, &fg.read_fds, &fg.write_fds, &fg.exc_fds, &fg.max_fd
);
//printf("curl msfd %d %d\n", curlMErr, fg.max_fd);
}
void HTTP_OP_SET::got_select(FDSET_GROUP&, double timeout) {
int iNumMsg;
HTTP_OP* hop = NULL;
//bool time_passed = false;
CURLMsg *pcurlMsg = NULL;
int iRunning = 0; // curl flags for max # of fds & # running queries
CURLMcode curlMErr;
CURLcode curlErr;
// get the data waiting for transfer in or out
// use timeout value so that we don't hog CPU in this loop
//
while (1) {
curlMErr = curl_multi_perform(g_curlMulti, &iRunning);
if (curlMErr != CURLM_CALL_MULTI_PERFORM) break;
if (dtime() - gstate.now > timeout) break;
}
// read messages from curl that may have come in from the above loop
//
while (1) {
pcurlMsg = curl_multi_info_read(g_curlMulti, &iNumMsg);
if (!pcurlMsg) break;
// if we have a msg, then somebody finished
// can check also with pcurlMsg->msg == CURLMSG_DONE
//
hop = lookup_curl(pcurlMsg->easy_handle);
if (!hop) continue;
// we have a message from one of our http_ops
// get the response code for this request
//
curlErr = curl_easy_getinfo(hop->curlEasy,
CURLINFO_RESPONSE_CODE, &hop->response
);
// CURLINFO_LONG+25 is a workaround for a bug in the gcc version
// included with Mac OS X 10.3.9
//
curlErr = curl_easy_getinfo(hop->curlEasy,
(CURLINFO)(CURLINFO_LONG+25) /*CURLINFO_OS_ERRNO*/, &hop->connect_error
);
// update byte counts and transfer speed
//
if (hop->want_download) {
bytes_down += hop->bytes_xferred;
curlErr = curl_easy_getinfo(hop->curlEasy,
CURLINFO_SPEED_DOWNLOAD, &hop->xfer_speed
);
}
if (hop->want_upload) {
bytes_up += hop->bytes_xferred;
curlErr = curl_easy_getinfo(hop->curlEasy,
CURLINFO_SPEED_UPLOAD, &hop->xfer_speed
);
}
// if proxy/socks server uses authentication and its not set yet,
// get what last transfer used
if (hop->auth_flag && !hop->auth_type) {
curlErr = curl_easy_getinfo(hop->curlEasy,
CURLINFO_PROXYAUTH_AVAIL, &hop->auth_type);
}
// the op is done if curl_multi_msg_read gave us a msg for this http_op
//
hop->http_op_state = HTTP_STATE_DONE;
hop->CurlResult = pcurlMsg->data.result;
if (hop->CurlResult == CURLE_OK) {
if ((hop->response/100)*100 == HTTP_STATUS_OK) {
hop->http_op_retval = 0;
} else if ((hop->response/100)*100 == HTTP_STATUS_CONTINUE) {
continue;
} else {
// Got a response from server but its not OK or CONTINUE,
// so save response with error message to display later.
//
if (hop->response >= 400) {
strcpy(hop->error_msg, boincerror(hop->response));
} else {
sprintf(hop->error_msg, "HTTP error %ld", hop->response);
}
switch (hop->response) {
case HTTP_STATUS_NOT_FOUND:
hop->http_op_retval = ERR_FILE_NOT_FOUND;
break;
default:
hop->http_op_retval = ERR_HTTP_ERROR;
}
}
net_status.need_physical_connection = false;
} else {
strcpy(hop->error_msg, curl_easy_strerror(hop->CurlResult));
switch(hop->CurlResult) {
case CURLE_COULDNT_RESOLVE_HOST:
hop->http_op_retval = ERR_GETHOSTBYNAME;
break;
case CURLE_COULDNT_CONNECT:
hop->http_op_retval = ERR_CONNECT;
break;
default:
hop->http_op_retval = ERR_HTTP_ERROR;
}
net_status.got_http_error();
if (log_flags.http_debug) {
msg_printf(NULL, MSG_ERROR, "[http_debug] HTTP error: %s", hop->error_msg);
}
}
if (!hop->http_op_retval && hop->http_op_type == HTTP_OP_POST2) {
// for a successfully completed request on a "post2" --
// read in the temp file into req1 memory
//
fclose(hop->fileOut);
double dSize = 0.0f;
file_size(hop->outfile, dSize);
hop->fileOut = boinc_fopen(hop->outfile, "rb");
if (!hop->fileOut) { // ack, can't open back up!
hop->response = 1;
// flag as a bad response for a possible retry later
} else {
fseek(hop->fileOut, 0, SEEK_SET);
// CMC Note: req1 is a pointer to "header" which is 4096
memset(hop->req1, 0, 4096);
fread(hop->req1, 1, (size_t) dSize, hop->fileOut);
}
}
// close files and "sockets" (i.e. libcurl handles)
//
hop->close_file();
hop->close_socket();
// finally remove the tmpfile if not explicitly set
//
if (hop->bTempOutfile) {
boinc_delete_file(hop->outfile);
}
}
}
// Return the HTTP_OP object with given Curl object
//
HTTP_OP* HTTP_OP_SET::lookup_curl(CURL* pcurl) {
for (unsigned int i=0; i<http_ops.size(); i++) {
if (http_ops[i]->curlEasy == pcurl) {
return http_ops[i];
}
}
return 0;
}
// Update the transfer speed for this HTTP_OP
// called on every I/O
//
void HTTP_OP::update_speed() {
double delta_t = dtime() - start_time;
if (delta_t > 0) {
xfer_speed = (bytes_xferred-start_bytes_xferred) / delta_t;
}
}
void HTTP_OP::set_speed_limit(bool is_upload, double bytes_sec) {
#if LIBCURL_VERSION_NUM >= 0x070f05
CURLcode cc = CURLE_OK;
curl_off_t bs = (curl_off_t)bytes_sec;
if (is_upload) {
cc = curl_easy_setopt(curlEasy, CURLOPT_MAX_SEND_SPEED_LARGE, bs);
} else {
cc = curl_easy_setopt(curlEasy, CURLOPT_MAX_RECV_SPEED_LARGE, bs);
}
if (cc && log_flags.http_debug) {
msg_printf(NULL, MSG_ERROR, "[http_debug] Curl error in set_speed_limit(): %s", curl_easy_strerror(cc));
}
#endif
}
const char *BOINC_RCSID_57f273bb60 = "$Id$";