Motivation:
For protocols like QUIC using UDP_SEGMENT (GSO) can help to reduce the
overhead quite a bit. We should support it.
Modifications:
- Add a SegmentedDatagramPacket which can be used to use UDP_SEGMENT
- Add unit test
Result:
Be able to make use of UDP_SEGMENT
Motivation:
To make it possible to experiment with alternative buffer implementations, we need a way to abstract away the concrete buffers used throughout most of the Netty pipelines, while still having a common currency for doing IO in the end.
Modification:
- Introduce an ByteBufConvertible interface, that allow arbitrary objects to convert themselves into ByteBuf objects.
- Every place in the code, where we did an instanceof check for ByteBuf, we now do an instanceof check for ByteBufConvertible.
- ByteBuf itself implements ByteBufConvertible, and returns itself from the asByteBuf method.
Result:
It is now possible to use Netty with alternative buffer implementations, as long as they can be converted to ByteBuf.
This has been verified elsewhere, with an alternative buffer implementation.
Support TCP Fast Open for clients and make SslHandler take advantage
Motivation:
- TCP Fast Open allow us to send a small amount of data along side the initial SYN packet when establishing a TCP connection.
- The TLS Client Hello packet is small enough to fit in there, and is also idempotent (another requirement for using TCP Fast Open), so if we can save a round-trip when establishing TLS connections when using TFO.
Modification:
- Add support for client-side TCP Fast Open for Epoll, and also lowers the Linux kernel version requirements to 3.6.
- When adding the SslHandler to a pipeline, if TCP Fast Open is enabled for the channel (and the channel is not already active) then start the handshake early by writing it to the outbound buffer.
- An important detail to note here, is that the outbound buffer is not flushed at this point, like it would for normal handshakes. The flushing happens later as part of establishing the TCP connection.
Result:
- It is now possible for clients (on epoll) to open connections with TCP Fast Open.
- The SslHandler automatically detects when this is the case, and now send its Client Hello message as part of the initial data in the TCP Fast Open flow when available, saving a round-trip when establishing TLS connections.
Co-authored-by: Colin Godsey <crgodsey@gmail.com>
Motivation:
File.createTempFile(String, String)` will create a temporary file in the system temporary directory if the 'java.io.tmpdir'. The permissions on that file utilize the umask. In a majority of cases, this means that the file that java creates has the permissions: `-rw-r--r--`, thus, any other local user on that system can read the contents of that file.
This can be a security concern if any sensitive data is stored in this file.
This was reported by Jonathan Leitschuh <jonathan.leitschuh@gmail.com> as a security problem.
Modifications:
Use Files.createTempFile(...) which will use safe-defaults when running on java 7 and later. If running on java 6 there isnt much we can do, which is fair enough as java 6 shouldnt be considered "safe" anyway.
Result:
Create temporary files with sane permissions by default.
Allow and skip null handlers when adding a vararg list of handlers
Motivation
Allowing null handlers allows for more convenient idioms in
conditionally adding handlers, e.g.,
ch.pipeline().addLast(
new FooHandler(),
condition ? new BarHandler() : null,
new BazHandler()
);
Modifications
* Change addFirst(..) and addLast(..) to skip null handlers, rather than
break or short-circuit.
* Add new unit tests.
Result
* Makes addFirst(..) and addLast(..) behavior more consistent
* Resolves https://github.com/netty/netty/issues/10728
Motivation:
We need to ensure we copy the attributes and options when bootstrap the channel as otherwise we may change the underlying Entry.
This is similar to what was reported in https://github.com/netty/netty-incubator-codec-quic/issues/152.
Modifications:
- Do a copy and re-use methods
- Add unit tests
Result:
Don't affect attributes / options of channels that are already bootstrapped
Motivation:
https in xmlns URIs does not work and will let the maven release plugin fail:
```
[INFO] ------------------------------------------------------------------------
[INFO] BUILD FAILURE
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 1.779 s
[INFO] Finished at: 2020-11-10T07:45:21Z
[INFO] ------------------------------------------------------------------------
[ERROR] Failed to execute goal org.apache.maven.plugins:maven-release-plugin:2.5.3:prepare (default-cli) on project netty-parent: Execution default-cli of goal org.apache.maven.plugins:maven-release-plugin:2.5.3:prepare failed: The namespace xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" could not be added as a namespace to "project": The namespace prefix "xsi" collides with an additional namespace declared by the element -> [Help 1]
[ERROR]
```
See also https://issues.apache.org/jira/browse/HBASE-24014.
Modifications:
Use http for xmlns
Result:
Be able to use maven release plugin
Motivation:
`DefaultChannelId` uses reflection to access the JMX runtime. This needs some extra config for GraalVM.
Modification:
Add config for GraalVM
Result:
Works when using GraalVM native image
Motivation:
During the last few month we did develop an io_uring based transport which shows very promising performance numbers. To give it more time to bake we will develop it outside of netty in an "incubator" module which will make it clear to users what to expect and also allow us to seperate its release cycle. While the implementation of it is very self contained there are few small adjustments that need to be made in netty itself to allow us to reuse code.
Modifications:
- AbstractChannel: Add method which can be used when a write fails and remove final from one method
- IovArray: Allow to create an IovArray from a ByteBuf instance
- FileDescriptor: Allow to reuse mark close logic via sub-class
Result:
Be able to reuse netty core classes in io_uring incubator repository
Motivation:
We should preferable always release the message before we notify the promise. Thhis has a few advantages:
- Release memory as soon as possible
- Listeners observe the "more correct" reference count
Modifications:
Release message before fail the promises
Result:
Faster releasing of resources. This came up in https://github.com/netty/netty/issues/10723
Motivation:
Since GraalVM version 19.3.0, instances of java.net.InetAddress (and its subclasses Inet4Address and Inet6Address) are not allowed in native image heap (that is, they cannot be stored in static fields of classes initialized at build time or be reachable through static fields of such classes). When building a native image, it makes sense to initialize at build time as many classes as possible.
But some fields of some classes in Netty (for example, NetUtil.LOCALHOST4) contain InetAddress instances. If a program is using code path that makes it possible to reach such fields at build time initialization, it becomes impossible to build a native image initializing core Netty classes initialized at runtime. An example of such a program is a client that uses netty-dns.
Modifications:
- Add netty-testsuite-native-image-client Maven module to test that such an example program can be built after the corresponding fixes
- Add native-image.properties to resolver-dns module to move initialization of some classes to runtime (some of them are parsing configuration during initialization, so it makes no sense to initialize them at build time; for others, it's needed to avoid InetAddress reachability at build time)
- Add substitutions for NetUtil.LOCALHOST4, NetUtil.LOCALHOST6 and NetUtil.LOCALHOST to overcome the InetAddress-related prohibition
- Extract some initialization code from NetUtil to NetUtilInitializations to allow it to be used by the substitutions
Result:
A client program using netty-dns with --initialize-at-build-time=io.netty builds successfully
Motivation:
Thread.stop() works by producing a ThreadDeath error in the target thread. EventLoops swallow all Throwables, which makes them effectively unkillable. This is effectively a memory leak, for our application. Beside this we should also just regrow all `Error` as there is almost no way to recover.
Modification:
Edit the EventLoops that swallow Throwables to instead rethrow Error.
Result:
`EventLoop` can crash if `Error` is thrown
Motivation:
HTTP is a plaintext protocol which means that someone may be able
to eavesdrop the data. To prevent this, HTTPS should be used whenever
possible. However, maintaining using https:// in all URLs may be
difficult. The nohttp tool can help here. The tool scans all the files
in a repository and reports where http:// is used.
Modifications:
- Added nohttp (via checkstyle) into the build process.
- Suppressed findings for the websites
that don't support HTTPS or that are not reachable
Result:
- Prevent using HTTP in the future.
- Encourage users to use HTTPS when they follow the links they found in
the code.
Motivation:
DuplexChannel allow for half-closure, we should have a special config interface for it as well.
Modifications:
Add DuplexChannelConfig which allows to configure half-closure.
Result:
More consistent types
Motivation:
junit deprecated Assert.assertThat(...)
Modifications:
Use MatcherAssert.assertThat(...) as replacement for deprecated method
Result:
Less deprecation warnings
Motivation:
LGTM reports multiple issues. They need to be triaged,
and real ones should be fixed.
Modifications:
- Fixed multiple issues reported by LGTM, such as redundant conditions,
resource leaks, typos, possible integer overflows.
- Suppressed false-positives.
- Added a few testcases.
Result:
Fixed several possible issues, get rid of false alarms in the LGTM report.
Motivation:
We wish to use Unsafe as little as possible, and Java 8 allows us
to take some short-cuts or play some tricks with generics,
for the purpose of working around having to declare all checked
exceptions. Ideally all checked exceptions would be declared, but
the code base is not ready for that yet.
Modification:
The call to UNSAFE.throwException has been removed, so when we need
that feature, we instead use the generic exception trick.
In may cases, Java 8 allows us to throw Throwable directly. This
happens in cases where no exception is declared to be thrown in a
scope.
Finally, some warnings have also been fixed, and some imports have
been reorganised and cleaned up while I was modifying the files
anyway.
Result:
We no longer use Unsafe for throwing any exceptions.
Motivation:
We should use an initial buffer size with is >= 1500 (which is a common setting for MTU) to reduce the need for memory copies when a new connection is established. This is especially interesting when SSL / TLS comes into the mix.
This was ported from swiftnio:
https://github.com/apple/swift-nio/pull/1641
Modifications:
Increase the initial size from 1024 to 2048.
Result:
Possible less memory copies on new connections
Motivation:
Creating exceptions is expensive so we should only do so if really needed.
Modifications:
Only create the ConnectTimeoutException if we really need it.
Result:
Less overhead
Motivation:
In some benchmarks closing the Channel attributes to a lot of overhead due the call of fillInStackTrace(). We should reduce this overhead.
Modifications:
- Create a StacklessClosedChannelException and use it to reduce overhead.
- Only call ChannelOutboundBuffer.failFlushed(...) when there was a flushed message at all.
Result:
Less performance overhead when closing the Channel
Motivation:
I was working on the transport part in Netty (ofc, solving a major issue) and I found this typo so thought to fix it.
Modification:
Fixed Typo
Result:
No more confusion between `us` and `use`.
Motivation:
Noticed we had some unused non-public classes.
There is no reason to keep these around.
Modification:
Remove unused non-public classes.
Result:
Less code to worry about.
Motivation:
When we were using the netty http protocol, OOM occurred, this problem has been in 4.1.51.Final Fix [# 10424](https://github.com/netty/netty/issues/10424), even if OOM is up, the service will still receive new connection events, will occur again OOM and eventually cause the connection not to be released.
code `byteBuf = allocHandle.allocate(allocator);`
Modification:
I fail to create buffer when I try to receive new data, i determine if it is OOM then the close read event releases the connection.
```java
if (close || cause instanceof OutOfMemoryError || cause instanceof IOException) {
closeOnRead(pipeline);
}
```
Result:
Fixes # [10434](https://github.com/netty/netty/issues/10434).
Motivation:
When a switch statement is used we should always define a `default:` so we don't introduce bugs due fall-through.
Modification:
Add missing `default:`s
Result:
Less error-prone code
Motivation:
AbstractCoalescingBufferQueue had a bug which could lead to an empty queue while still report bytes left. This was due the fact that we decremented the pending bytes before draining the queue one-by-one. The problem here is that while the queue is drained we may notify the promise which may add again buffers to the queue for which we never decrement the bytes while we drain these
Modifications:
- Decrement the pending bytes every time we drain a buffer from the queue
- Add unit tests
Result:
Fixes https://github.com/netty/netty/issues/10286
Motivation:
In DefaultChannelHandlerContext we had some code where we tried to guard against endless loops caused by exceptions thrown by exceptionCaught(...) that would trigger exceptionCaught again. This code was proplematic for two reasons:
- It is quite expensive as we need to compare all the stacks
- We may end up not notify another handlers exceptionCaught(...) if in our exeuction stack we triggered actions that will cause an exceptionCaught somewhere else in the pipeline
Modifications:
- Just remove the detection code as we already handle everything correctly when we invoke exceptionCaught(...)
- Add unit tests
Result:
Ensure we always notify correctly and also fixes performance issue reported as https://github.com/netty/netty/issues/10165
Motivation:
Parameter maxCount needs the unit test.
Modifications:
1. Change the conditional statement to avoid the ineffective maxCount (enhance the robustness of the code merely).
2. Add the unit test for maxCount.
Result:
Enable this parameter to be tested.
Motivation:
For whatever reason IntelliJ produces some testfailures when using mock(ChannelHandler.class). In this case it does not "correctly" detect that the @Skip annotation should not be taken into effect.
Modifications:
Use spy(...) to ensure we correctly invoke methods
Result:
No more tests failures when running via IntelliJ.
Motivation:
As we have java8 as a minimum target we can use MethodHandles. We should do so when we expect to have a method called multiple times.
Modifications:
- Replace usage of reflection with MethodHandles where it makes sense
- Remove some code which was there to support java < 8
Result:
Faster code
Motivation:
In general, we will close the debug log in a product environment. However, logging without external level check may still affect performance as varargs will need to allocate an array.
Modification:
Add log level check simply before logging.
Result:
Improve performance slightly in a product environment.
Motivation:
At the moment the next / prev references are not set to "null" in the DefaultChannelHandlerContext once the ChannelHandler is removed. This is bad as it basically let users still use the ChannelHandlerContext of a ChannelHandler after it is removed and may produce very suprising behaviour.
Modifications:
- Fail if someone tries to use the ChannelHandlerContext once the ChannelHandler was removed (for outbound operations fail the promise, for inbound fire the error through the ChannelPipeline)
- Fix some handlers to ensure we not use the ChannelHandlerContext after the handler was removed
- Adjust DefaultChannelPipeline / DefaultChannelHandlerContext to fixes races with removal / replacement of handlers
Result:
Cleanup behaviour and make it more predictable for pipeline modifications
Motivation:
We should not include the number of ServerChannel that are part of the DefaultChannelGroup when specify the initial size of the LinkedHashMap
Modifications:
Only use the number of the non ServerChannel
Result:
Reduce memory-footprint
Motivation:
Current pipeline handler replace tests are replacing handler with same name.
we need test that test handler can renamed, old handlers are removed from pipeline.
Modification:
There is coverage missing related to renaming of handlers
Result:
Adds missing tests
Co-authored-by: phani254 <phani254@yahoo.com>
Motivation:
https://github.com/netty/netty/pull/9458 changed how we handled ChannelOptions internally to use a ConcurrentHashMap. This unfortunally had the side-effect that the ordering may be affected and not stable anymore. Here the problem is that sometimes we do validation based on two different ChannelOptions (for example we validate high and low watermarks against each other). Thus even if the user specified the options in the same order we may fail to configure them.
Modifications:
- Use again a LinkedHashMap to preserve order
- Add unit test
Result:
Apply ChannelOptions in correct and expected order
Motivation:
Due a bug in DefaultChannelPipeline we did not call handlerRemoved(...) if the handlers was removed via removeLast() or removeFirst()
Modifications:
- Correctly implement removeLast() / removeFirst()
- Add unit tests
Result:
handlerRemoved(...) always called on removal
Motivation:
Because we now tie and EventLoop to a Channel from the beginning its possible that we schedule something on it before the registration with the Selector is done. This can lead to AssertionErrors in the current form of the code like:
java.lang.AssertionError
at io.netty.channel.nio.AbstractNioChannel.selectionKey(AbstractNioChannel.java:109)
at io.netty.channel.nio.AbstractNioChannel$AbstractNioUnsafe.isFlushPending(AbstractNioChannel.java:344)
at io.netty.channel.nio.AbstractNioChannel$AbstractNioUnsafe.flush0(AbstractNioChannel.java:332)
at io.netty.channel.AbstractChannel$AbstractUnsafe.flush(AbstractChannel.java:855)
at io.netty.channel.DefaultChannelPipeline$HeadHandler.flush(DefaultChannelPipeline.java:1253)
at io.netty.channel.DefaultChannelHandlerContext.invokeFlush(DefaultChannelHandlerContext.java:747)
at io.netty.channel.DefaultChannelHandlerContext.access$1300(DefaultChannelHandlerContext.java:38)
at io.netty.channel.DefaultChannelHandlerContext$WriteAndFlushTask.write(DefaultChannelHandlerContext.java:1144)
at io.netty.channel.DefaultChannelHandlerContext$AbstractWriteTask.run(DefaultChannelHandlerContext.java:1055)
at io.netty.util.concurrent.SingleThreadEventExecutor.runAllTasks(SingleThreadEventExecutor.java:360)
at io.netty.channel.SingleThreadEventLoop.run(SingleThreadEventLoop.java:200)
at io.netty.util.concurrent.SingleThreadEventExecutor.lambda$doStartThread$4(SingleThreadEventExecutor.java:855)
at io.netty.util.internal.ThreadExecutorMap$2.run(ThreadExecutorMap.java:75)
at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
at java.lang.Thread.run(Thread.java:748)
Modifications:
Remove assert and add a few null checks.
Result:
No more java.lang.AssertionError.
Motivation:
In next major version of netty users should use ChannelHandler everywhere. We should ensure we do the same
Modifications:
Replace usage of deprecated classes / interfaces with ChannelHandler
Result:
Use non-deprecated code
Motivation:
97361fa2c8 replace synchronized with ConcurrentHashMap in *Bootstrap classes but missed to do the same for the Http2 variant.
Modifications:
- Use ConcurrentHashMap
- Simplify code in *Bootstrap classes
Result:
Less contention
Motivation:
In netty 5.x we changed to have all pipeline operations be executed on the EventLoop so there is no need to have an atomic operation involved anymore to update the handler state.
Modifications:
Remove atomic usage to handle the handler state
Result:
Simpler code and less overhead
Motivation:
In most cases, we want to use MultithreadEventLoopGroup without setting thread numbers but thread name only. We should simplify this for the user
Modifications:
Add a new constructor
Result:
User can only set ThreadFactory / Executor without setting the thread number to 0:
Motivation
Per javadoc in 4.1.x SimpleChannelInboundHandler:
"Please keep in mind that channelRead0(ChannelHandlerContext, I) will be
renamed to messageReceived(ChannelHandlerContext, I) in 5.0."
Modifications
Rename aforementioned method and all references/overrides.
Result
Method is renamed.
Motivation:
At the moment we directly extend the Recycler base class in our code which makes it hard to experiment with different Object pool implementation. It would be nice to be able to switch from one to another by using a system property in the future. This would also allow to more easily test things like https://github.com/netty/netty/pull/8052.
Modifications:
- Introduce ObjectPool class with static method that we now use internally to obtain an ObjectPool implementation.
- Wrap the Recycler into an ObjectPool and return it for now
Result:
Preparation for different ObjectPool implementations
Motivation
The current event loop shutdown logic is quite fragile and in the
epoll/NIO cases relies on the default 1 second wait/select timeout that
applies when there are no scheduled tasks. Without this default timeout
the shutdown would hang indefinitely.
The timeout only takes effect in this case because queued scheduled
tasks are first cancelled in
SingleThreadEventExecutor#confirmShutdown(), but I _think_ even this
isn't robust, since the main task queue is subsequently serviced which
could result in some new scheduled task being queued with much later
deadline.
It also means shutdowns are unnecessarily delayed by up to 1 second.
Modifications
- Add/extend unit tests to expose the issue
- Adjust SingleThreadEventExecutor shutdown and confirmShutdown methods
to explicitly add no-op tasks to the taskQueue so that the subsequent
event loop iteration doesn't enter blocking wait (as looks like was
originally intended)
Results
Faster and more robust shutdown of event loops, allows removal of the default wait timeout.
This is a port of https://github.com/netty/netty/pull/9616
Motivation:
On JDK > 9 Netty uses Unsafe to write two internal JDK fields: sun.nio.ch.SelectorImp.selectedKeys and sun.nio.ch.SelectorImpl.publicSelectedKeys. This is done in transport/src/main/java/io/netty/channel/nio/NioEventLoop.java:225, in openSelector() method. The GraalVM analysis cannot do the Unsafe registration automatically because the object field offset computation is hidden behind two layers of calls.
Modifications:
This PR updates the Netty GraalVM configuration by registering those fields for unsafe access.
Result:
Improved support for Netty on GraalVM with JDK > 9.
Motivation:
Due a bug we did not always correctly calculate the next buffer size in AdaptiveRecvByteBufAllocator.
Modification:
Fix calculation and add unit test
Result:
Correct calculation is always used.
Motiviation:
A lot of reentrancy bugs and cycles can happen because the DefaultPromise will notify the FutureListener directly when completely in the calling Thread if the Thread is the EventExecutor Thread. To reduce the risk of this we should always notify the listeners via the EventExecutor which basically means that we will put a task into the taskqueue of the EventExecutor and pick it up for execution after the setSuccess / setFailure methods complete the promise.
Modifications:
- Always notify via the EventExecutor
- Adjust test to ensure we correctly account for this
- Adjust tests that use the EmbeddedChannel to ensure we execute the scheduled work.
Result:
Reentrancy bugs related to the FutureListeners cant happen anymore.
