* kryptos at freenetproject.org <kryptos at freenetproject.org> [2007-08-27
23:44:24]:
> Author: kryptos
> Date: 2007-08-27 23:44:24 +0000 (Mon, 27 Aug 2007)
> New Revision: 14895
>
> Modified:
> branches/freenet-jfk/src/freenet/node/FNPPacketMangler.java
> Log:
> Error in JFK message caching corrected thanks to nextgens
>
> Modified: branches/freenet-jfk/src/freenet/node/FNPPacketMangler.java
> ===================================================================
> --- branches/freenet-jfk/src/freenet/node/FNPPacketMangler.java
> 2007-08-27 23:37:38 UTC (rev 14894)
> +++ branches/freenet-jfk/src/freenet/node/FNPPacketMangler.java
> 2007-08-27 23:44:24 UTC (rev 14895)
> @@ -672,15 +672,21 @@
> * This would result in increased processing on the
> Responder side->CPU exhaustion attacks
> */
> byte[] cacheKey=processMessageAuth(pn);
> + Object result;
> //All recent messages 3 and 4 are cached
> if(phase==2){
> - message3Cache.put(cacheKey,data);
> - //if duplicate message3; send corresponding message4
> -
> if(data.toString().equalsIgnoreCase(message3Cache.get(cacheKey).toString())){
> - sendMessage4Packet(1,2,3,data,pn,replyTo);
> - return true;
> -
> + synchronized(message3Cache) {
> + result = message3Cache.get(cacheKey);
> }
> + if(result != null) {
> + synchronized(message3Cache) {
> + message3Cache.put(cacheKey,data);
> + }
> + // We don't want to keep the lock while sending
> + sendMessage4Packet(1,2,3,data,pn,replyTo);
> + return true;
> + }
> +
Hmmm, that's not what you want !
"
synchronized(message3Cache) {
result = message3Cache.get(cacheKey);
if(result == null) {
message3Cache.put(cacheKey,methodComputingData());
}
}
// We don't want to keep the lock while sending
sendMessage4Packet(1,2,3,result,pn,replyTo);
"
Is closer but not ideal. Keep in mind that the whole point of caching
the message is to avoid "replays"/DoSes. You shouldn't spend any time in
ProcessMessage3 before calling that method (ie: do the expensive stuffs
in methodComputingData())
By the way I suggest you keep a static instance of HMAC so that you
don't spare time going through constructors/destructors in
processMessageAuth(PeerNode).
Hope this helps,
NextGen$
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20070828/583dfea1/attachment.pgp>
