/* * Copyright (c) 2013, 2017 EBay Software Foundation and others. All rights reserved. * * This program and the accompanying materials are made available under the * terms of the Eclipse Public License v1.0 which accompanies this distribution, * and is available at http://www.eclipse.org/legal/epl-v10.html */ package org.opendaylight.ovsdb.lib.jsonrpc; import com.fasterxml.jackson.core.JsonEncoding; import com.fasterxml.jackson.core.JsonFactory; import com.fasterxml.jackson.core.JsonParser; import com.fasterxml.jackson.core.io.IOContext; import com.fasterxml.jackson.core.json.ByteSourceJsonBootstrapper; import com.fasterxml.jackson.core.util.BufferRecycler; import com.fasterxml.jackson.databind.JsonNode; import com.fasterxml.jackson.databind.MappingJsonFactory; import io.netty.buffer.ByteBuf; import io.netty.buffer.ByteBufInputStream; import io.netty.channel.ChannelHandlerContext; import io.netty.handler.codec.ByteToMessageDecoder; import java.io.IOException; import java.util.List; import org.opendaylight.ovsdb.lib.error.InvalidEncodingException; import org.slf4j.Logger; import org.slf4j.LoggerFactory; /** * JSON RPC 1.0 compatible decoder capable of decoding JSON messages from a TCP stream. * The stream is framed first by inspecting the json for valid end marker (left curly) * and is passed to a Json parser (jackson) for converting into an object model. * * <p>There are no JSON parsers that I am aware of that does non blocking parsing. * This approach avoids having to run json parser over and over again on the entire * stream waiting for input. Parser is invoked only when we know of a full JSON message * in the stream. */ public class JsonRpcDecoder extends ByteToMessageDecoder { private static final Logger LOG = LoggerFactory.getLogger(JsonRpcDecoder.class); private final int maxFrameLength; //Indicates if the frame limit warning was issued private boolean maxFrameLimitWasReached = false; private final JsonFactory jacksonJsonFactory = new MappingJsonFactory(); private final IOContext jacksonIOContext = new IOContext(new BufferRecycler(), null, false); // context for the previously read incomplete records private int lastRecordBytes = 0; private int leftCurlies = 0; private int rightCurlies = 0; private boolean inS = false; private int recordsRead; public JsonRpcDecoder(int maxFrameLength) { this.maxFrameLength = maxFrameLength; } @Override protected void decode(ChannelHandlerContext ctx, ByteBuf buf, List<Object> out) throws Exception { LOG.trace("readable bytes {}, records read {}, incomplete record bytes {}", buf.readableBytes(), recordsRead, lastRecordBytes); if (lastRecordBytes == 0) { if (buf.readableBytes() < 4) { return; //wait for more data } skipSpaces(buf); byte[] buff = new byte[4]; buf.getBytes(buf.readerIndex(), buff); ByteSourceJsonBootstrapper strapper = new ByteSourceJsonBootstrapper(jacksonIOContext, buff, 0, 4); JsonEncoding jsonEncoding = strapper.detectEncoding(); if (!JsonEncoding.UTF8.equals(jsonEncoding)) { throw new InvalidEncodingException(jsonEncoding.getJavaName(), "currently only UTF-8 is supported"); } } int index = lastRecordBytes + buf.readerIndex(); for (; index < buf.writerIndex(); index++) { switch (buf.getByte(index)) { case '{': if (!inS) { leftCurlies++; } break; case '}': if (!inS) { rightCurlies++; } break; case '"': if (buf.getByte(index - 1) != '\\') { inS = !inS; } break; default: break; } if (leftCurlies != 0 && leftCurlies == rightCurlies && !inS) { ByteBuf slice = buf.readSlice(1 + index - buf.readerIndex()); JsonParser jp = jacksonJsonFactory.createParser(new ByteBufInputStream(slice)); JsonNode root = jp.readValueAsTree(); out.add(root); leftCurlies = 0; rightCurlies = 0; lastRecordBytes = 0; recordsRead++; break; } /* * Changing this limit to being a warning, we do not wish to "break" in scale environment * and currently this limits the ovs of having only around 50 ports defined... * I do acknowledge the fast that this might be risky in case of huge amount of strings * in which the controller can crash with an OOM, however seems that we need a really huge * ovs to reach that limit. */ //We do not want to issue a log message on every extent of the buffer //hence logging only once if (index - buf.readerIndex() >= maxFrameLength && !maxFrameLimitWasReached) { maxFrameLimitWasReached = true; LOG.warn("***** OVSDB Frame limit of {} bytes has been reached! *****", this.maxFrameLength); } } // end of stream, save the incomplete record index to avoid reexamining the whole on next run if (index >= buf.writerIndex()) { lastRecordBytes = buf.readableBytes(); } } public int getRecordsRead() { return recordsRead; } private static void skipSpaces(ByteBuf byteBuf) throws IOException { while (byteBuf.isReadable()) { int ch = byteBuf.getByte(byteBuf.readerIndex()) & 0xFF; if (!(ch == ' ' || ch == '\r' || ch == '\n' || ch == '\t')) { return; } else { byteBuf.readByte(); //move the read index } } } }