/**
* Callback, invoked on each successfully parsed response header key/value pair.
*/
bool HealthMonitor::onMessageHeader(const BufferRef& name, const BufferRef& value)
{
TRACE("onResponseHeader(name:%s, value:%s)", name.str().c_str(), value.str().c_str());
if (x0::iequals(name, "Status")) {
int status = value.ref(0, value.find(' ')).toInt();
responseCode_ = static_cast<x0::HttpStatus>(status);
}
return true;
}
bool HttpConnection::onMessageHeader(const BufferRef& name, const BufferRef& value)
{
if (request_->isFinished()) {
// this can happen when the request has failed some checks and thus,
// a client error message has been sent already.
// we need to "parse" the remaining content anyways.
TRACE("onMessageHeader() skip \"%s\": \"%s\"", name.str().c_str(), value.str().c_str());
return true;
}
TRACE("onMessageHeader() \"%s\": \"%s\"", name.str().c_str(), value.str().c_str());
if (iequals(name, "Host")) {
auto i = value.find(':');
if (i != BufferRef::npos)
request_->hostname = value.ref(0, i);
else
request_->hostname = value;
TRACE(" -- hostname set to \"%s\"", request_->hostname.str().c_str());
} else if (iequals(name, "Connection")) {
if (iequals(value, "close"))
setShouldKeepAlive(false);
else if (iequals(value, "keep-alive"))
setShouldKeepAlive(true);
}
// limit the size of a single request header
if (name.size() + value.size() > worker().server().maxRequestHeaderSize()) {
TRACE("header too long. got %ld / %ld", name.size() + value.size(), worker().server().maxRequestHeaderSize());
request_->status = HttpError::RequestEntityTooLarge;
request_->finish();
return false;
}
// limit the number of request headers
if (request_->requestHeaders.size() > worker().server().maxRequestHeaderCount()) {
TRACE("header count exceeded. got %ld / %ld", request_->requestHeaders.size(), worker().server().maxRequestHeaderCount());
request_->status = HttpError::RequestEntityTooLarge;
request_->finish();
return false;
}
request_->requestHeaders.push_back(HttpRequestHeader(name, value));
return true;
}
/** processes a message-chunk.
*
* \param chunk the chunk of bytes to process
*
* \return number of bytes actually parsed and processed
*/
std::size_t HttpMessageProcessor::process(const BufferRef& chunk, size_t* out_nparsed)
{
/*
* CR = 0x0D
* LF = 0x0A
* SP = 0x20
* HT = 0x09
*
* CRLF = CR LF
* LWS = [CRLF] 1*( SP | HT )
*
* HTTP-message = Request | Response
*
* generic-message = start-line
* *(message-header CRLF)
* CRLF
* [ message-body ]
*
* start-line = Request-Line | Status-Line
*
* Request-Line = Method SP Request-URI SP HTTP-Version CRLF
*
* Method = "OPTIONS" | "GET" | "HEAD"
* | "POST" | "PUT" | "DELETE"
* | "TRACE" | "CONNECT"
* | extension-method
*
* Request-URI = "*" | absoluteURI | abs_path | authority
*
* extension-method = token
*
* Status-Line = HTTP-Version SP Status-Code SP Reason-Phrase CRLF
*
* HTTP-Version = "HTTP" "/" 1*DIGIT "." 1*DIGIT
* Status-Code = 3*DIGIT
* Reason-Phrase = *<TEXT, excluding CR, LF>
*
* absoluteURI = "http://" [user ':'******'@'] hostname [abs_path] [qury]
* abs_path = "/" *CHAR
* authority = ...
* token = 1*<any CHAR except CTLs or seperators>
* separator = "(" | ")" | "<" | ">" | "@"
* | "," | ";" | ":" | "\" | <">
* | "/" | "[" | "]" | "?" | "="
* | "{" | "}" | SP | HT
*
* message-header = field-name ":" [ field-value ]
* field-name = token
* field-value = *( field-content | LWS )
* field-content = <the OCTETs making up the field-value
* and consisting of either *TEXT or combinations
* of token, separators, and quoted-string>
*
* message-body = entity-body
* | <entity-body encoded as per Transfer-Encoding>
*/
const char* i = chunk.cbegin();
const char* e = chunk.cend();
const size_t initialOutOffset = out_nparsed ? *out_nparsed : 0;
size_t result = initialOutOffset;
size_t* nparsed = out_nparsed ? out_nparsed : &result;
//TRACE(2, "process(curState:%s): size: %ld: '%s'", state_str(), chunk.size(), chunk.str().c_str());
TRACE(2, "process(curState:%s): size: %ld", state_str(), chunk.size());
#if 0
switch (state_) {
case CONTENT: // fixed size content
if (!passContent(chunk, nparsed))
goto done;
i += *nparsed;
break;
case CONTENT_ENDLESS: // endless-sized content (until stream end)
{
*nparsed += chunk.size();
bool rv = filters_.empty()
? onMessageContent(chunk)
: onMessageContent(filters_.process(chunk));
goto done;
}
default:
break;
}
#endif
while (i != e) {
#if !defined(XZERO_NDEBUG)
if (std::isprint(*i)) {
TRACE(3, "parse: %4ld, 0x%02X (%c), %s", *nparsed, *i, *i, state_str());
} else {
TRACE(3, "parse: %4ld, 0x%02X, %s", *nparsed, *i, state_str());
}
#endif
switch (state_) {
case MESSAGE_BEGIN:
contentLength_ = -1;
switch (mode_) {
case REQUEST:
state_ = REQUEST_LINE_BEGIN;
versionMajor_ = 0;
versionMinor_ = 0;
break;
case RESPONSE:
state_ = STATUS_LINE_BEGIN;
code_ = 0;
versionMajor_ = 0;
versionMinor_ = 0;
break;
case MESSAGE:
state_ = HEADER_NAME_BEGIN;
// an internet message has no special top-line,
// so we just invoke the callback right away
if (!onMessageBegin())
goto done;
break;
}
break;
case REQUEST_LINE_BEGIN:
if (isToken(*i)) {
state_ = REQUEST_METHOD;
method_ = chunk.ref(*nparsed - initialOutOffset, 1);
++*nparsed;
++i;
} else {
state_ = SYNTAX_ERROR;
}
break;
case REQUEST_METHOD:
if (*i == SP) {
state_ = REQUEST_ENTITY_BEGIN;
++*nparsed;
++i;
} else if (!isToken(*i)) {
state_ = SYNTAX_ERROR;
} else {
method_.shr();
++*nparsed;
++i;
}
break;
case REQUEST_ENTITY_BEGIN:
if (std::isprint(*i)) {
entity_ = chunk.ref(*nparsed - initialOutOffset, 1);
state_ = REQUEST_ENTITY;
++*nparsed;
++i;
}
else {
state_ = SYNTAX_ERROR;
}
break;
case REQUEST_ENTITY:
if (*i == SP) {
state_ = REQUEST_PROTOCOL_BEGIN;
++*nparsed;
++i;
} else if (std::isprint(*i)) {
entity_.shr();
++*nparsed;
++i;
} else
state_ = SYNTAX_ERROR;
break;
case REQUEST_PROTOCOL_BEGIN:
if (*i != 'H') {
state_ = SYNTAX_ERROR;
} else {
state_ = REQUEST_PROTOCOL_T1;
++*nparsed;
++i;
}
break;
case REQUEST_PROTOCOL_T1:
if (*i != 'T') {
state_ = SYNTAX_ERROR;
} else {
state_ = REQUEST_PROTOCOL_T2;
++*nparsed;
++i;
}
break;
case REQUEST_PROTOCOL_T2:
if (*i != 'T') {
state_ = SYNTAX_ERROR;
} else {
state_ = REQUEST_PROTOCOL_P;
++*nparsed;
++i;
}
break;
case REQUEST_PROTOCOL_P:
if (*i != 'P') {
state_ = SYNTAX_ERROR;
} else {
state_ = REQUEST_PROTOCOL_SLASH;
++*nparsed;
++i;
}
break;
case REQUEST_PROTOCOL_SLASH:
if (*i != '/') {
state_ = SYNTAX_ERROR;
} else {
state_ = REQUEST_PROTOCOL_VERSION_MAJOR;
++*nparsed;
++i;
}
break;
case REQUEST_PROTOCOL_VERSION_MAJOR:
if (*i == '.') {
state_ = REQUEST_PROTOCOL_VERSION_MINOR;
++*nparsed;
++i;
}
else if (!std::isdigit(*i)) {
state_ = SYNTAX_ERROR;
} else {
versionMajor_ = versionMajor_ * 10 + *i - '0';
++*nparsed;
++i;
}
break;
case REQUEST_PROTOCOL_VERSION_MINOR:
if (*i == CR) {
state_ = REQUEST_LINE_LF;
++*nparsed;
++i;
}
#if defined(X0_HTTP_SUPPORT_SHORT_LF)
else if (*i == LF) {
state_ = HEADER_NAME_BEGIN;
++*nparsed;
++i;
TRACE(2, "request-line: method=%s, entity=%s, vmaj=%d, vmin=%d",
method_.str().c_str(), entity_.str().c_str(), versionMajor_, versionMinor_);
if (!onMessageBegin(method_, entity_, versionMajor_, versionMinor_)) {
goto done;
}
}
#endif
else if (!std::isdigit(*i)) {
state_ = SYNTAX_ERROR;
} else {
versionMinor_ = versionMinor_ * 10 + *i - '0';
++*nparsed;
++i;
}
break;
case REQUEST_LINE_LF:
if (*i == LF) {
state_ = HEADER_NAME_BEGIN;
++*nparsed;
++i;
TRACE(2, "request-line: method=%s, entity=%s, vmaj=%d, vmin=%d",
method_.str().c_str(), entity_.str().c_str(), versionMajor_, versionMinor_);
if (!onMessageBegin(method_, entity_, versionMajor_, versionMinor_)) {
goto done;
}
}
else
state_ = SYNTAX_ERROR;
break;
case STATUS_LINE_BEGIN:
case STATUS_PROTOCOL_BEGIN:
if (*i != 'H') {
state_ = SYNTAX_ERROR;
} else {
state_ = STATUS_PROTOCOL_T1;
++*nparsed;
++i;
}
break;
case STATUS_PROTOCOL_T1:
if (*i != 'T') {
state_ = SYNTAX_ERROR;
} else {
state_ = STATUS_PROTOCOL_T2;
++*nparsed;
++i;
}
break;
case STATUS_PROTOCOL_T2:
if (*i != 'T') {
state_ = SYNTAX_ERROR;
} else {
state_ = STATUS_PROTOCOL_P;
++*nparsed;
++i;
}
break;
case STATUS_PROTOCOL_P:
if (*i != 'P') {
state_ = SYNTAX_ERROR;
} else {
state_ = STATUS_PROTOCOL_SLASH;
++*nparsed;
++i;
}
break;
case STATUS_PROTOCOL_SLASH:
if (*i != '/') {
state_ = SYNTAX_ERROR;
} else {
state_ = STATUS_PROTOCOL_VERSION_MAJOR;
++*nparsed;
++i;
}
break;
case STATUS_PROTOCOL_VERSION_MAJOR:
if (*i == '.') {
state_ = STATUS_PROTOCOL_VERSION_MINOR;
++*nparsed;
++i;
} else if (!std::isdigit(*i)) {
state_ = SYNTAX_ERROR;
} else {
versionMajor_ = versionMajor_ * 10 + *i - '0';
++*nparsed;
++i;
}
break;
case STATUS_PROTOCOL_VERSION_MINOR:
if (*i == SP) {
state_ = STATUS_CODE_BEGIN;
++*nparsed;
++i;
} else if (!std::isdigit(*i)) {
state_ = SYNTAX_ERROR;
} else {
versionMinor_ = versionMinor_ * 10 + *i - '0';
++*nparsed;
++i;
}
break;
case STATUS_CODE_BEGIN:
if (!std::isdigit(*i)) {
code_ = SYNTAX_ERROR;
break;
}
state_ = STATUS_CODE;
/* fall through */
case STATUS_CODE:
if (std::isdigit(*i)) {
code_ = code_ * 10 + *i - '0';
++*nparsed;
++i;
} else if (*i == SP) {
state_ = STATUS_MESSAGE_BEGIN;
++*nparsed;
++i;
} else if (*i == CR) { // no Status-Message passed
state_ = STATUS_MESSAGE_LF;
++*nparsed;
++i;
} else {
state_ = SYNTAX_ERROR;
}
break;
case STATUS_MESSAGE_BEGIN:
if (isText(*i)) {
state_ = STATUS_MESSAGE;
message_ = chunk.ref(*nparsed - initialOutOffset, 1);
++*nparsed;
++i;
}
else
state_ = SYNTAX_ERROR;
break;
case STATUS_MESSAGE:
if (isText(*i) && *i != CR && *i != LF) {
message_.shr();
++*nparsed;
++i;
} else if (*i == CR) {
state_ = STATUS_MESSAGE_LF;
++*nparsed;
++i;
} else {
state_ = SYNTAX_ERROR;
}
break;
case STATUS_MESSAGE_LF:
if (*i == LF) {
state_ = HEADER_NAME_BEGIN;
++*nparsed;
++i;
//TRACE(2, "status-line: HTTP/%d.%d, code=%d, message=%s", versionMajor_, versionMinor_, code_, message_.str().c_str());
if (!onMessageBegin(versionMajor_, versionMinor_, code_, message_)) {
goto done;
}
} else
state_ = SYNTAX_ERROR;
break;
case HEADER_NAME_BEGIN:
if (isToken(*i)) {
name_ = chunk.ref(*nparsed - initialOutOffset, 1);
state_ = HEADER_NAME;
++*nparsed;
++i;
} else if (*i == CR) {
state_ = HEADER_END_LF;
++*nparsed;
++i;
}
#if defined(X0_HTTP_SUPPORT_SHORT_LF)
else if (*i == LF)
state_ = HEADER_END_LF;
#endif
else
state_ = SYNTAX_ERROR;
break;
case HEADER_NAME:
if (isToken(*i)) {
name_.shr();
++*nparsed;
++i;
} else if (*i == ':') {
state_ = LWS_BEGIN;
lwsNext_ = HEADER_VALUE_BEGIN;
lwsNull_ = HEADER_VALUE_END; // only (CR LF) parsed, assume empty value & go on with next header
++*nparsed;
++i;
} else if (*i == CR) {
state_ = LWS_LF;
lwsNext_ = HEADER_COLON;
lwsNull_ = SYNTAX_ERROR;
++*nparsed;
++i;
} else
state_ = SYNTAX_ERROR;
break;
case HEADER_COLON:
if (*i == ':') {
state_ = LWS_BEGIN;
lwsNext_ = HEADER_VALUE_BEGIN;
lwsNull_ = HEADER_VALUE_END;
++*nparsed;
++i;
} else
state_ = SYNTAX_ERROR;
break;
case LWS_BEGIN:
if (*i == CR) {
state_ = LWS_LF;
++*nparsed;
++i;
#if defined(X0_HTTP_SUPPORT_SHORT_LF)
} else if (*i == LF) {
state_ = LWS_SP_HT_BEGIN;
++*nparsed;
++i;
#endif
} else if (*i == SP || *i == HT) {
state_ = LWS_SP_HT;
++*nparsed;
++i;
} else if (std::isprint(*i)) {
state_ = lwsNext_;
} else
state_ = SYNTAX_ERROR;
break;
case LWS_LF:
if (*i == LF) {
state_ = LWS_SP_HT_BEGIN;
++*nparsed;
++i;
} else
state_ = SYNTAX_ERROR;
break;
case LWS_SP_HT_BEGIN:
if (*i == SP || *i == HT) {
if (!value_.empty())
value_.shr(3); // CR LF (SP | HT)
state_ = LWS_SP_HT;
++*nparsed;
++i;
} else {
// only (CF LF) parsed so far and no 1*(SP | HT) found.
state_ = lwsNull_;
// XXX no nparsed/i-update
}
break;
case LWS_SP_HT:
if (*i == SP || *i == HT) {
if (!value_.empty())
value_.shr();
++*nparsed;
++i;
} else
state_ = lwsNext_;
break;
case HEADER_VALUE_BEGIN:
if (isText(*i)) {
value_ = chunk.ref(*nparsed - initialOutOffset, 1);
++*nparsed;
++i;
state_ = HEADER_VALUE;
} else if (*i == CR) {
state_ = HEADER_VALUE_LF;
++*nparsed;
++i;
#if defined(X0_HTTP_SUPPORT_SHORT_LF)
} else if (*i == LF) {
state_ = HEADER_VALUE_END;
++*nparsed;
++i;
#endif
} else {
state_ = SYNTAX_ERROR;
}
break;
case HEADER_VALUE:
if (*i == CR) {
state_ = LWS_LF;
lwsNext_ = HEADER_VALUE;
lwsNull_ = HEADER_VALUE_END;
++*nparsed;
++i;
}
#if defined(X0_HTTP_SUPPORT_SHORT_LF)
else if (*i == LF) {
state_ = LWS_SP_HT_BEGIN;
lwsNext_ = HEADER_VALUE;
lwsNull_ = HEADER_VALUE_END;
++*nparsed;
++i;
}
#endif
else if (isText(*i)) {
value_.shr();
++*nparsed;
++i;
} else
state_ = SYNTAX_ERROR;
break;
case HEADER_VALUE_LF:
if (*i == LF) {
state_ = HEADER_VALUE_END;
++*nparsed;
++i;
} else {
state_ = SYNTAX_ERROR;
}
break;
case HEADER_VALUE_END: {
TRACE(2, "header: name='%s', value='%s'", name_.str().c_str(), value_.str().c_str());
if (iequals(name_, "Content-Length")) {
contentLength_ = value_.toInt();
TRACE(2, "set content length to: %ld", contentLength_);
} else if (iequals(name_, "Transfer-Encoding")) {
if (iequals(value_, "chunked")) {
chunked_ = true;
}
}
bool rv = onMessageHeader(name_, value_);
name_.clear();
value_.clear();
// continue with the next header
state_ = HEADER_NAME_BEGIN;
if (!rv) {
goto done;
}
break;
}
case HEADER_END_LF:
if (*i == LF) {
if (isContentExpected())
state_ = CONTENT_BEGIN;
else
state_ = MESSAGE_BEGIN;
++*nparsed;
++i;
if (!onMessageHeaderEnd()) {
TRACE(2, "messageHeaderEnd returned false. returning `Aborted`-state");
goto done;
}
if (!isContentExpected() && !onMessageEnd()) {
goto done;
}
} else {
state_ = SYNTAX_ERROR;
}
break;
case CONTENT_BEGIN:
if (chunked_)
state_ = CONTENT_CHUNK_SIZE_BEGIN;
else if (contentLength_ >= 0)
state_ = CONTENT;
else
state_ = CONTENT_ENDLESS;
break;
case CONTENT_ENDLESS: {
// body w/o content-length (allowed in simple MESSAGE types only)
BufferRef c(chunk.ref(*nparsed - initialOutOffset));
//TRACE(2, "prepared content-chunk (%ld bytes): %s", c.size(), c.str().c_str());
*nparsed += c.size();
i += c.size();
bool rv = filters_.empty()
? onMessageContent(c)
: onMessageContent(filters_.process(c).ref());
if (!rv)
goto done;
break;
}
case CONTENT: {
// fixed size content length
std::size_t offset = *nparsed - initialOutOffset;
std::size_t chunkSize = std::min(static_cast<size_t>(contentLength_), chunk.size() - offset);
contentLength_ -= chunkSize;
*nparsed += chunkSize;
i += chunkSize;
bool rv = filters_.empty()
? onMessageContent(chunk.ref(offset, chunkSize))
: onMessageContent(filters_.process(chunk.ref(offset, chunkSize)).ref());
if (contentLength_ == 0)
state_ = MESSAGE_BEGIN;
if (!rv)
goto done;
if (state_ == MESSAGE_BEGIN && !onMessageEnd())
goto done;
break;
}
case CONTENT_CHUNK_SIZE_BEGIN:
if (!std::isxdigit(*i)) {
state_ = SYNTAX_ERROR;
break;
}
state_ = CONTENT_CHUNK_SIZE;
contentLength_ = 0;
/* fall through */
case CONTENT_CHUNK_SIZE:
if (*i == CR) {
state_ = CONTENT_CHUNK_LF1;
++*nparsed;
++i;
} else if (*i >= '0' && *i <= '9') {
contentLength_ = contentLength_ * 16 + *i - '0';
++*nparsed;
++i;
} else if (*i >= 'a' && *i <= 'f') {
contentLength_ = contentLength_ * 16 + 10 + *i - 'a';
++*nparsed;
++i;
} else if (*i >= 'A' && *i <= 'F') {
contentLength_ = contentLength_ * 16 + 10 + *i - 'A';
++*nparsed;
++i;
} else {
state_ = SYNTAX_ERROR;
}
break;
case CONTENT_CHUNK_LF1:
if (*i != LF) {
state_ = SYNTAX_ERROR;
} else {
//TRACE(2, "content_length: %ld", contentLength_);
if (contentLength_ != 0)
state_ = CONTENT_CHUNK_BODY;
else
state_ = CONTENT_CHUNK_CR3;
++*nparsed;
++i;
}
break;
case CONTENT_CHUNK_BODY:
if (contentLength_) {
std::size_t offset = *nparsed - initialOutOffset;
std::size_t chunkSize = std::min(static_cast<size_t>(contentLength_), chunk.size() - offset);
contentLength_ -= chunkSize;
*nparsed += chunkSize;
i += chunkSize;
bool rv = filters_.empty()
? onMessageContent(chunk.ref(offset, chunkSize))
: onMessageContent(filters_.process(chunk.ref(offset, chunkSize)).ref());
if (!rv) {
goto done;
}
} else if (*i == CR) {
state_ = CONTENT_CHUNK_LF2;
++*nparsed;
++i;
}
break;
case CONTENT_CHUNK_LF2:
if (*i != LF) {
state_ = SYNTAX_ERROR;
} else {
state_ = CONTENT_CHUNK_SIZE;
++*nparsed;
++i;
}
break;
case CONTENT_CHUNK_CR3:
if (*i != CR) {
state_ = SYNTAX_ERROR;
} else {
state_ = CONTENT_CHUNK_LF3;
++*nparsed;
++i;
}
break;
case CONTENT_CHUNK_LF3:
if (*i != LF) {
state_ = SYNTAX_ERROR;
} else {
++*nparsed;
++i;
if (!onMessageEnd())
goto done;
state_ = MESSAGE_BEGIN;
}
break;
case SYNTAX_ERROR: {
#if !defined(XZERO_NDEBUG)
TRACE(1, "parse: syntax error");
if (std::isprint(*i)) {
TRACE(1, "parse: syntax error at nparsed: %ld, character: '%c'", *nparsed, *i);
} else {
TRACE(1, "parse: syntax error at nparsed: %ld, character: 0x%02X", *nparsed, *i);
}
chunk.dump("request chunk (at syntax error)");
#endif
goto done;
}
default:
#if !defined(XZERO_NDEBUG)
TRACE(1, "parse: unknown state %i", state_);
if (std::isprint(*i)) {
TRACE(1, "parse: internal error at nparsed: %ld, character: '%c'", *nparsed, *i);
} else {
TRACE(1, "parse: internal error at nparsed: %ld, character: 0x%02X", *nparsed, *i);
}
Buffer::dump(chunk.data(), chunk.size(), "request chunk (at unknown state)");
#endif
goto done;
}
}
// we've reached the end of the chunk
if (state_ == CONTENT_BEGIN) {
// we've just parsed all headers but no body yet.
if (contentLength_ < 0 && !chunked_ && mode_ != MESSAGE) {
// and there's no body to come
if (!onMessageEnd())
goto done;
// subsequent calls to process() parse next request(s).
state_ = MESSAGE_BEGIN;
}
}
done:
return *nparsed - initialOutOffset;
}