// returns false on bad mime
bool HttpMime::parse ( char *mime , long mimeLen , Url *url ) {
// reset locUrl to 0
m_locUrl.reset();
// return if we have no valid complete mime
if ( mimeLen == 0 ) return false;
// status is on first line
m_status = -1;
// skip HTTP/x.x till we hit a space
char *p = mime;
char *pend = mime + mimeLen;
while ( p < pend && !is_wspace_a(*p) ) p++;
// then skip over spaces
while ( p < pend && is_wspace_a(*p) ) p++;
// return false on a problem
if ( p == pend ) return false;
// then read in the http status
m_status = atol2 ( p , pend - p );
// if no Content-Type: mime field was provided, assume html
m_contentType = CT_HTML;
// assume default charset
m_charset = NULL;
m_charsetLen = 0;
// set contentLen, lastModifiedDate, m_cookie
p = mime;
while ( p < pend ) {
// compute the length of the string starting at p and ending
// at a \n or \r
long len = 0;
while ( &p[len] < pend && p[len]!='\n' && p[len]!='\r' ) len++;
// . if we could not find a \n or \r there was an error
// . MIMEs must always end in \n or \r
if ( &p[len] >= pend ) return false;
// . stick a NULL at the end of the line
// . overwrites \n or \r TEMPORARILY
char c = p [ len ];
p [ len ] = '\0';
// parse out some meaningful data
if ( strncasecmp ( p , "Content-Length:" ,15) == 0 ) {
m_contentLengthPos = p + 15;
m_contentLen = atol( m_contentLengthPos);
}
else if ( strncasecmp ( p , "Last-Modified:" ,14) == 0 ) {
m_lastModifiedDate=atotime(p+14);
// do not let them exceed current time for purposes
// of sorting by date using datedb (see Msg16.cpp)
time_t now = time(NULL);
if (m_lastModifiedDate > now) m_lastModifiedDate = now;
}
else if ( strncasecmp ( p , "Content-Type:" ,13) == 0 )
m_contentType = getContentTypePrivate ( p + 13 );
else if ( strncasecmp ( p , "Set-Cookie: " ,11) == 0 ) {
m_cookie = p + 11;
m_cookieLen = gbstrlen ( p + 11 );
}
else if ( strncasecmp ( p , "Location:" , 9) == 0 ) {
// point to it
char *tt = p + 9;
// skip if space
if ( *tt == ' ' ) tt++;
if ( *tt == ' ' ) tt++;
// at least set this for Msg13.cpp to use
m_locationField = tt;
m_locationFieldLen = gbstrlen(tt);
// . we don't add the "www." because of slashdot.com
// . we skip initial spaces in this Url::set() routine
if(url)
m_locUrl.set ( url, p + 9, len - 9,
false/*addWWW?*/);
}
else if ( strncasecmp ( p , "Content-Encoding:", 17) == 0 ) {
//only support gzip now, it doesn't seem like servers
//implement the other types much
m_contentEncodingPos = p+17;
if(strstr(m_contentEncodingPos, "gzip")) {
m_contentEncoding = ET_GZIP;
}
else if(strstr(m_contentEncodingPos, "deflate")) {
//zlib's compression
m_contentEncoding = ET_DEFLATE;
}
}
//else if ( strncasecmp ( p, "Cookie:", 7) == 0 )
// log (LOG_INFO, "mime: Got Cookie = %s", (p+7));
// re-insert the character that we replaced with a '\0'
p [ len ] = c;
// go to next line
p += len;
// skip over the cruft at the end of this line
while ( p < pend && ( *p=='\r' || *p=='\n' ) ) p++;
}
return true;
}
// . when the Conf::m_proxyIps parm is updated we call this to rebuild
// s_iptab, our table of SpiderProxy instances, which has the proxies and
// their performance statistics.
// . we try to maintain stats of ip/ports that did NOT change when rebuilding.
bool buildProxyTable ( ) {
// scan the NEW list of proxy ip/port pairs in g_conf
char *p = g_conf.m_proxyIps.getBufStart();
HashTableX tmptab;
tmptab.set(8,0,16,NULL,0,false,"tmptab");
// scan the user inputted space-separated list of ip:ports
// (optional username:password@ip:port)
for ( ; *p ; ) {
// skip white space
if ( is_wspace_a(*p) ) { p++; continue; }
// skip http://
if ( strncasecmp(p,"http://",7) == 0 ) { p += 7; continue; }
// scan in an ip:port
char *s = p; char *portStr = NULL;
int32_t dc = 0, pc = 0, gc = 0, bc = 0;
const char *msg;
char *usernamePwd = NULL;
int32_t usernamePwdLen = 0;
char *ipStart = p;
// scan all characters until we hit \0 or another whitespace
for ( ; *s && !is_wspace_a(*s); s++) {
if ( *s == '@' ) {
// must be username:pwd
if ( pc != 1 ) {
msg = "bad username:password";
goto hadError;
}
usernamePwd = p;
usernamePwdLen = s - p;
if ( usernamePwdLen >= MAXUSERNAMEPWD-2 ) {
msg = "username:password too long";
goto hadError;
}
dc = 0;
gc = 0;
bc = 0;
pc = 0;
portStr = NULL;
ipStart = s+1;
continue;
}
if ( *s == '.' ) { dc++; continue; }
if ( *s == ':' ) { portStr=s; pc++; continue; }
if ( is_digit(*s) ) { gc++; continue; }
bc++;
continue;
}
// ensure it is a legit ip:port combo
msg = NULL;
if ( gc < 4 )
msg = "not enough digits for an ip";
if ( pc > 1 )
msg = "too many colons";
if ( dc != 3 )
msg = "need 3 dots for an ip address";
if ( bc )
msg = "got illegal char in ip:port listing";
if ( msg ) {
hadError:
char c = *s;
*s = '\0';
log("buf: %s for %s",msg,p);
*s = c;
return false;
}
// convert it
int32_t iplen = s - ipStart;
if ( portStr ) iplen = portStr - ipStart;
int32_t ip = atoip(ipStart,iplen);
// another sanity check
if ( ip == 0 || ip == -1 ) {
log("spider: got bad proxy ip for %s",p);
return false;
}
// and the port default is 80
int32_t port = 80;
if ( portStr ) port = atol2(portStr+1,s-portStr-1);
if ( port < 0 || port > 65535 ) {
log("spider: got bad proxy port for %s",p);
return false;
}
// . we got a legit ip:port
// . see if already in our table
uint64_t ipKey = (uint32_t)ip;
ipKey <<= 16;
ipKey |= (uint16_t)(port & 0xffff);
// also store into tmptable to see what we need to remove
tmptab.addKey(&ipKey);
// see if in table
int32_t islot = s_iptab.getSlot( &ipKey);
// advance p
p = s;
// if in there, keep it as is
if ( islot >= 0 ) continue;
// otherwise add new entry
SpiderProxy newThing;
memset ( &newThing , 0 , sizeof(SpiderProxy));
newThing.m_ip = ip;
newThing.m_port = port;
newThing.m_lastDownloadTookMS = -1;
newThing.m_lastSuccessfulTestMS = -1;
gbmemcpy(newThing.m_usernamePwd,usernamePwd,usernamePwdLen);
// ensure it is NULL terminated
newThing.m_usernamePwd[usernamePwdLen] = '\0';
if ( ! s_iptab.addKey ( &ipKey, &newThing ) )
return false;
}
redo:
int32_t removed = 0;
// scan all SpiderProxies in tmptab
for ( int32_t i = 0 ; i < s_iptab.getNumSlots() ; i++ ) {
// skip empty buckets in hashtable s_iptab
if ( ! s_iptab.m_flags[i] ) continue;
// get the key
int64_t key = *(int64_t *)s_iptab.getKeyFromSlot(i);
// must also exist in tmptab, otherwise it got removed by user
if ( tmptab.isInTable ( &key ) ) continue;
// skip if not in table
if ( s_iptab.getSlot ( &key ) < 0 ) {
log("sproxy: iptable hashing messed up");
continue;
}
// shoot, it got removed. not in the new list of ip:ports
s_iptab.removeKey ( &key );
removed++;
// hashtable is messed up now, start over
//goto redo;
}
if ( removed ) goto redo;
return true;
}
// returns false on bad mime
bool HttpMime::parse(char *mime, int32_t mimeLen, Url *url) {
#ifdef _VALGRIND_
VALGRIND_CHECK_MEM_IS_DEFINED(mime,mimeLen);
#endif
// reset locUrl to 0
m_locUrl.reset();
// return if we have no valid complete mime
if (mimeLen == 0) {
return false;
}
// status is on first line
m_status = -1;
// skip HTTP/x.x till we hit a space
char *p = mime;
char *pend = mime + mimeLen;
while (p < pend && !is_wspace_a(*p)) p++;
// then skip over spaces
while (p < pend && is_wspace_a(*p)) p++;
// return false on a problem
if (p == pend) return false;
// then read in the http status
m_status = atol2(p, pend - p);
// if no Content-Type: mime field was provided, assume html
m_contentType = CT_HTML;
// assume default charset
m_charset = NULL;
m_charsetLen = 0;
// skip over first line
getNextLine();
while (getNextLine()) {
const char *field = NULL;
size_t fieldLen = 0;
if (getField(&field, &fieldLen)) {
if (parseContentEncoding(field, fieldLen)) {
continue;
}
if (parseContentLength(field, fieldLen)) {
continue;
}
if (parseContentType(field, fieldLen)) {
continue;
}
if (parseLocation(field, fieldLen, url)) {
continue;
}
if (parseSetCookie(field, fieldLen)) {
continue;
}
// add parsing of other header here
}
}
return true;
}
