Daniel Chaffelson created NIFI-15690:
----------------------------------------
Summary: CacheRequestDecoder.readHeader() leaks ByteBuf - should
use skipBytes instead of readBytes
Key: NIFI-15690
URL: https://issues.apache.org/jira/browse/NIFI-15690
Project: Apache NiFi
Issue Type: Bug
Affects Versions: 2.7.2, 1.28.1
Reporter: Daniel Chaffelson
The readHeader() method in CacheRequestDecoder allocates a new ByteBuf that is
never released:
{code:java}
private void readHeader(final ByteBuf byteBuf, final SocketAddress
remoteAddress) {
if (byteBuf.readableBytes() >= HEADER_LENGTH) {
byteBuf.readBytes(HEADER_LENGTH); // Creates new ByteBuf, never
released
headerReceived.getAndSet(true);
log.debug("Header Received [{}]", remoteAddress);
}
}
{code}
In Netty, {{ByteBuf.readBytes(int)}} allocates a *new* ByteBuf containing the
extracted bytes. Since the return value is discarded, this buffer is never
released, triggering Netty's ResourceLeakDetector:
{noformat}
LEAK: ByteBuf.release() was not called before it's garbage-collected.
...
at
org.apache.nifi.distributed.cache.server.codec.CacheRequestDecoder.readHeader(CacheRequestDecoder.java:238)
{noformat}
*Fix*: Replace {{readBytes(HEADER_LENGTH)}} with {{skipBytes(HEADER_LENGTH)}}
since the header content is not used.
*Affected versions*: 1.17.0+ and all 2.x (introduced in NIFI-9805)
*Impact*: Low - 4 bytes leaked per new cache client connection. Affects flows
using DistributedMapCacheServer, DistributedSetCacheServer, Wait, Notify,
DetectDuplicate processors.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
