*/
package org.opendaylight.controller.netconf.nettyutil.handler;
-import java.util.List;
-
-import org.opendaylight.controller.netconf.api.NetconfMessage;
-import org.opendaylight.controller.netconf.util.xml.XmlUtil;
-import org.slf4j.Logger;
-import org.slf4j.LoggerFactory;
-
-import com.google.common.annotations.VisibleForTesting;
-
import io.netty.buffer.ByteBuf;
import io.netty.buffer.ByteBufInputStream;
import io.netty.buffer.ByteBufUtil;
+import io.netty.buffer.Unpooled;
import io.netty.channel.ChannelHandlerContext;
import io.netty.handler.codec.ByteToMessageDecoder;
+import java.io.IOException;
+import java.util.List;
+import org.opendaylight.controller.netconf.api.NetconfMessage;
+import org.opendaylight.controller.netconf.util.xml.XmlUtil;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import org.xml.sax.SAXException;
public final class NetconfXMLToMessageDecoder extends ByteToMessageDecoder {
private static final Logger LOG = LoggerFactory.getLogger(NetconfXMLToMessageDecoder.class);
@Override
- @VisibleForTesting
- public void decode(ChannelHandlerContext ctx, ByteBuf in, List<Object> out) throws Exception {
+ public void decode(final ChannelHandlerContext ctx, final ByteBuf in, final List<Object> out) throws IOException, SAXException {
+ if (in.isReadable()) {
+ if (LOG.isTraceEnabled()) {
+ LOG.trace("Received to decode: {}", ByteBufUtil.hexDump(in));
+ }
+
+ /* According to the XML 1.0 specifications, when there is an XML declaration
+ * at the beginning of an XML document, it is invalid to have
+ * white spaces before that declaration (reminder: a XML declaration looks like:
+ * <?xml version="1.0" encoding="UTF-8"?>). In contrast, when there is no XML declaration,
+ * it is valid to have white spaces at the beginning of the document.
+ *
+ * When they send a NETCONF message, several NETCONF servers start with a new line (either
+ * LF or CRLF), presumably to improve readability in interactive sessions with a human being.
+ * Some NETCONF servers send an XML declaration, some others do not.
+ *
+ * If a server starts a NETCONF message with white spaces and follows with an XML
+ * declaration, XmlUtil.readXmlToDocument() will fail because this is invalid XML.
+ * But in the spirit of the "NETCONF over SSH" RFC 4742 and to improve interoperability, we want
+ * to accept those messages.
+ *
+ * To do this, the following code strips the leading bytes before the start of the XML messages.
+ */
- if (in.readableBytes() != 0) {
- LOG.trace("Received to decode: {}", ByteBufUtil.hexDump(in));
+ // Skip all leading whitespaces by moving the reader index to the first non whitespace character
+ while (in.isReadable()) {
+ if (!isWhitespace(in.readByte())) {
+ // return reader index to the first non whitespace character
+ in.readerIndex(in.readerIndex() - 1);
+ break;
+ }
+ }
+
+ // Warn about leading whitespaces
+ if (in.readerIndex() != 0 && LOG.isWarnEnabled()) {
+ final byte[] strippedBytes = new byte[in.readerIndex()];
+ in.getBytes(0, strippedBytes, 0, in.readerIndex());
+ LOG.warn("XML message with unwanted leading bytes detected. Discarded the {} leading byte(s): '{}'",
+ in.readerIndex(), ByteBufUtil.hexDump(Unpooled.wrappedBuffer(strippedBytes)));
+ }
+ }
+ if (in.isReadable()) {
out.add(new NetconfMessage(XmlUtil.readXmlToDocument(new ByteBufInputStream(in))));
} else {
LOG.debug("No more content in incoming buffer.");
}
}
+
+ /**
+ * Check whether a byte is whitespace/control character. Considered whitespace characters: <br/>
+ * SPACE, \t, \n, \v, \r, \f
+ *
+ * @param b byte to check
+ * @return true if the byte is a whitespace/control character
+ */
+ private static boolean isWhitespace(final byte b) {
+ return b <= 0x0d && b >= 0x09 || b == 0x20;
+ }
}