Motivation:
Exceptions generated from transport-native-epoll may include duplicate error string description or inconsistent usage of the method name in the string description.
Modifications:
- Ensure the method name from static exceptions and dynamic exceptions is of the same format
- Remove duplicate string rational from the exception messages
Result:
More consistent error messages with no duplicate error description.
Motivation:
We missed some stuff in 5728e0eb2c and so the build failed on java9
Modifications:
- Add extra cmdline args when needed
- skip the autobahntestsuite as jython not works with java9
- skip the osgi testsuite as the maven plugin not works with java9
Result:
Build finally passed on java9
Motivation:
EpollDatagramChannel uses getOption in the isActive method. getOption is backed by a relatively large conditional if/else if block and this conditional checking can be avoided in the epoll transport.
Modifications:
- Add EpollDatagramChannelConfig#getActiveOnOpen and use this in EpollDatagramChannel
Result:
Conditional checking due to getOption is removed from EpollDatagramChannel.
Motivation:
When the EPOLLRDHUP event is received we assume that the read side of the FD is no longer functional and force the input state to be shutdown. However if the channel is still active we should rely upon EPOLLIN and read to indicate there is no more data before we update the shutdown state. If we do not do this we may not read all pending data in the FD if the RecvByteBufAllocator doesn't want to consume it all in a single read operation.
Modifications:
- AbstractEpollChannel#epollRdHupReady() shouldn't force shutdown the input if the channel is active
Result:
All data can be read even if the RecvByteBufAllocator doesn't read it all in the current read loop.
Fixes https://github.com/netty/netty/issues/6303
Motivation:
We need to pass special arguments to run with jdk9 as otherwise some tests will not be able to run.
Modifications:
Allow to define extra arguments when running with jdk9
Result:
Tests pass with jdk9
Motivation:
EPOLL annotates some exceptions to provide the remote address, but the original exception is not preserved. This may make determining a root cause more difficult. The static EPOLL exceptions references the native method that failed, but does not provide a description of the actual error number. Without the description users have to know intimate details about the native calls and how they may fail to debug issues.
Modifications:
- annotated exceptions should preserve the original exception
- static exceptions should include the string description of the expected errno
Result:
EPOLL exceptions provide more context and are more useful to end users.
Motivation:
EpollRecvByteAllocatorHandle intends to override the meaning of "maybe more data to read" which is a concept also used in all existing implementations of RecvByteBufAllocator$Handle but the interface doesn't support overriding. Because the interfaces lack the ability to propagate this computation EpollRecvByteAllocatorHandle attempts to implement a heuristic on top of the delegate which may lead to reading when we shouldn't or not reading data.
Modifications:
- Create a new interface ExtendedRecvByteBufAllocator and ExtendedHandle which allows the "maybe more data to read" between interfaces
- Deprecate RecvByteBufAllocator and change all existing implementations to extend ExtendedRecvByteBufAllocator
- transport-native-epoll should require ExtendedRecvByteBufAllocator so the "maybe more data to read" can be propagated to the ExtendedHandle
Result:
Fixes https://github.com/netty/netty/issues/6303.
Motivation:
We should call Epoll.ensureAvailability() when init EpollEventLoopGroup to fail fast and with a proper exception.
Modifications:
Call Epoll.ensureAvailability() during EpollEventLoopGroup init.
Result:
Fail fast if epoll is not availability (for whatever reason).
Motivation:
EpollRecvByteAllocatorHandle will read unconditionally if EPOLLRDHUP has been received. However we can just treat this the same was we do as data maybe pending in ET mode, and let LT mode notify us if we haven't read all data.
Modifications:
- EpollRecvByteAllocatorHandle should not always force a read just because EPOLLRDHUP has been received, but just treated as an indicator that there maybe more data to read in ET mode
Result:
Fixes https://github.com/netty/netty/issues/6173.
Motivation:
In later Java8 versions our Atomic*FieldUpdater are slower then the JDK implementations so we should not use ours anymore. Even worse the JDK implementations provide for example an optimized version of addAndGet(...) using intrinsics which makes it a lot faster for this use-case.
Modifications:
- Remove methods that return our own Atomic*FieldUpdaters.
- Use the JDK implementations everywhere.
Result:
Faster code.
Motivation:
To keep our code clean we should fail the build when unused c code is detected.
Modifications:
- Add '-Wunused-variable' to build flags
Result:
Cleaner code.
Motivation:
When attempting to flamegraph netty w/ epoll it was noticed the stacks are lost going from
java to epoll lib.
Modifications:
added the -fno-omit-framepointer flag to compiler flags to ensure the fp is kept intact
Result:
Flamegraphs will now show native code in the same stack as java code using perf-java-flames
Motivation:
We missed to set active = true in EpollServerDomainSocketChannel.doBind(...) which also means that channelActive(...) was never triggered.
Modifications:
Correct set active = true in doBind(...)
Result:
EpollServerDomainSocketChannel is correctly set to active when bound.
Motivation:
I had a need to know the user credentials of a connected unix domain socket.
Modifications:
Added a class to encapsulate user credentials (UID, GID, and the PID).
Augemented the Socket class to provide the JNI native interface to return this new class
Augemented the c code to call getSockOpts passing <a href=http://man7.org/linux/man-pages/man7/socket.7.html>SO_PEERCRED</a>
Then surfaced the ability to get user credentials in the EpollDomainSocketChannel
Result:
The EpollDomainSocketChannel now has a the following function signature:
public PeerCredentials peerCredentials() throws IOException allowing a caller to get the UID, GID, and PID of the linux process
connected to the unix domain socket.
Motivation:
The previously generated manifest causes a parse exception when loaded into an Apache Felix OSGI container.
Modifications:
Fix parameter delimiter and unbalanced quotes in manifest entry. Suffixed with asterisk so the bundle is resolved on other architectures as well even if native libs won't be loaded.
Result:
Bundle will load properly in OSGI containers.
Motivation:
If an exception is thrown while processing the ready channels in the EventLoop we should still run all tasks as this may allow to recover. For example a OutOfMemoryError may be thrown and runAllTasks() will free up memory again. Beside this we should also ensure we always allow to shutdown even if an exception was thrown.
Modifications:
- Call runAllTasks() in a finally block
- Ensure shutdown is always handles.
Result:
More robust EventLoop implementations for NIO and Epoll.
Motivation:
At the moment only DefaultFileRegion is supported when using the native epoll transport.
Modification:
- Add support for any FileRegion implementation
- Add test case
Result:
Also custom FileRegion implementation are supported when using the epoll transport.
Motivition:
NIO throws ClosedChannelException when a user tries to call shutdown*() on a closed Channel. We should do the same for the EPOLL transport
Modification:
Throw ClosedChannelException when a user tries to call shutdown*() on a closed channel.
Result:
Consistent behavior.
Motivation:
Commit 2c1f17faa2 introduced a regression which could cause NotYetConnectedExceptions when handling RDHUP events.
Modifications:
Correct ignore NotYetConnectedException when handling RDHUP events.
Result:
No more regression.
Motivation:
The NIO transport used an IllegalStateException if a user tried to issue another connect(...) while the connect was still in process. For this case the JDK specified a ConnectPendingException which we should use. The same issues exists in the EPOLL transport. Beside this the EPOLL transport also does not throw the right exceptions for ENETUNREACH and EISCONN errno codes.
Modifications:
- Replace IllegalStateException with ConnectPendingException in NIO and EPOLL transport
- throw correct exceptions for ENETUNREACH and EISCONN in EPOLL transport
- Add test case
Result:
More correct error handling for connect attempts when using NIO and EPOLL transport
Motivation:
We need to ensure we also call fireChannelActive() if the Channel is directly closed in a ChannelFutureListener that is belongs to the promise for the connect. Otherwise we will see missing active events.
Modifications:
Ensure we always call fireChannelActive() if the Channel was active.
Result:
No missing events.
Motivation:
We should throw a NotYetConnectedException when ENOTCONN errno is set. This is also consistent with NIO.
Modification:
Throw correct exception and add test case
Result:
More correct and consistent behavior.
Motivation:
In 4.0 AbstractNioByteChannel has a default of 16 max messages per read. However in 4.1 that constraint was applied at the NioSocketChannel which is not equivalent. In 4.1 AbstractEpollStreamChannel also did not have the default of 16 max messages per read applied.
Modifications:
- Make Nio consistent with 4.0
- Make Epoll consistent with Nio
Result:
Nio and Epoll both have consistent ChannelMetadata and are consistent with 4.0.
Motivation:
The build generates a OSGi bundle with missing Bundle-NativeCode manifest entry.
Modifications:
Add missing manifest entry.
Result:
Be able to use transport-native-epoll in osgi container.
Motivation:
ECONNREFUSED can be a common type of exception when attempting to finish the connection process. Generating a new exception each time can be costly and quickly bloat memory usage.
Modifications:
- Expose ECONNREFUSED from JNI and cache this exception in Socket.finishConnect
Result:
ECONNREFUSED during finish connect doesn't create a new exception each time.
Motiviation:
Sometimes it is useful to allow to specify a custom strategy to handle rejected tasks. For example if someone tries to add tasks from outside the eventloop it may make sense to try to backoff and retries and so give the executor time to recover.
Modification:
Add RejectedEventExecutor interface and implementations and allow to inject it.
Result:
More flexible handling of executor overload.
Motivation:
To restrict the memory usage of a system it is sometimes needed to adjust the number of max pending tasks in the tasks queue.
Modifications:
- Add new constructors to modify the number of allowed pending tasks.
- Add system properties to configure the default values.
Result:
More flexible configuration.
Motivation:
We use pre-instantiated exceptions in various places for performance reasons. These exceptions don't include a stacktrace which makes it hard to know where the exception was thrown. This is especially true as we use the same exception type (for example ChannelClosedException) in different places. Setting some StackTraceElements will provide more context as to where these exceptions original and make debugging easier.
Modifications:
Set a generated StackTraceElement on these pre-instantiated exceptions which at least contains the origin class and method name. The filename and linenumber are specified as unkown (as stated in the javadocs of StackTraceElement).
Result:
Easier to find the origin of a pre-instantiated exception.
Motivation:
Unused methods create warnings on some C compilers. It may not be feasible to selectively turn them off.
Modifications:
Remove createInetSocketAddress as it is unused.
Result:
Less noisy compilation
Motivation:
JCTools supports both non-unsafe, unsafe versions of queues and JDK6 which allows us to shade the library in netty-common allowing it to stay "zero dependency".
Modifications:
- Remove copy paste JCTools code and shade the library (dependencies that are shaded should be removed from the <dependencies> section of the generated POM).
- Remove usage of OneTimeTask and remove it all together.
Result:
Less code to maintain and easier to update JCTools and less GC pressure as the queue implementation nt creates so much garbage
Motivation:
epoll_wait accepts a timeout argument which will specify the maximum amount of time the epoll_wait will wait for an event to occur. If the epoll_wait method returns for any reason that is not fatal (e.g. EINTR) the original timeout value is re-used. This does not honor the timeout interface contract and can lead to unbounded time in epoll_wait.
Modifications:
- The time taken by epoll_wait should be decremented before calling epoll_wait again, and if the remaining time is exhausted we should return 0 according to the epoll_wait interface docs http://man7.org/linux/man-pages/man2/epoll_wait.2.html
- link librt which is needed for some platforms to use clock_gettime
Result:
epoll_wait will wait for at most timeout ms according to the epoll_wait interface contract.
Motivation:
Sometimes it may be benefitially for an user to specify a custom algorithm when choose the next EventExecutor/EventLoop.
Modifications:
Allow to specify a custom EventExecutorChooseFactory that allows to customize algorithm.
Result:
More flexible api.
Motivation:
We used transfered in native code which is not correct spelling. It should be transferred.
Modifications:
Fix typo.
Result:
Less typos in source code.
Motivation:
Currenlty, netty-transport-native-epoll-*-linux-x86_64.jar is not packed as OSGi bundle
and thus not working in OSGi environment.
Modifications:
In netty-transport-native-epoll's pom.xml added configuration
to attach manifest to the jar with a native library.
In netty-common's pom.xml added configuration instruction (DynamicImport-Package)
to maven bnd plugin to make sure the native code is loaded from
netty-transport-native-epoll bundle.
Result:
The netty-transport-native-epoll-*-linux-x86_64.jar is a bundle (MANIFEST.MF attached)
and the inluced native library can be successfuly loaded in OSGi environment.
Fixing #5119
Motivation:
SingleThreadEventExecutor.pendingTasks() will call taskQueue.size() to get the number of pending tasks in the queue. This is not safe when using MpscLinkedQueue as size() is only allowed to be called by a single consumer.
Modifications:
Ensure size() is only called from the EventLoop.
Result:
No more livelock possible when call pendingTasks, no matter from which thread it is done.
Motivation:
The DuplexChannel is currently incomplete and only supports shutting down the output side of a channel. This interface should also support shutting down the input side of the channel.
Modifications:
- Add shutdownInput and shutdown methods to the DuplexChannel interface
- Remove state in NIO and OIO for tracking input being shutdown independent of the underlying transport's socket type. Tracking the state independently may lead to inconsistent state.
Result:
DuplexChannel supports shutting down the input side of the channel
Fixes https://github.com/netty/netty/issues/5175
Motivation:
If a task was submitted when wakenUp value was 1, the task didn't get a chance to produce wakeup event. So we need to check task queue again before calling epoll_wait. If we don't, the task might be pended until epoll_wait was timed out. It might be pended until idle timeout if IdleStateHandler existed in pipeline.
Modifications:
Execute epoll_wait in a non-blocking manner if there's a task submitted when wakenUp value was 1.
Result:
Every tasks in EpollEventLoop will not be pended.
Motivation:
EventExecutor.children uses generics in such a way that an entire colleciton must be cast to a specific type of object. This interface is not very flexible and is impossible to implement if the EventExecutor type must be wrapped. The current usage of this method also does not have any clear need within Netty. The Iterator interface allows for EventExecutor to be wrapped and forces the caller to make assumptions about types instead of building the assumptions into the interface.
Motivation:
- Remove EventExecutor.children and undeprecate the iterator() interface
Result:
EventExecutor interface has one less method and is easier to wrap.
Motivation:
When epoll datagram channel invokes sendmmsg0, _all_ of the messages go
on the wire with the address of the _last_ packet in the list.
Modifications:
An array of addresses equal to the length of the messages is allocated
on the stack to hold the address for each msg_hdr.msg_name.
Result:
Each message goes on the wire with the correct address.
Motivation:
NioEventLoopGroup supports constructors which take an executor but EpollEventLoopGroup does not. EPOLL should be consistent with NIO where ever possible.
Modifications:
- Add constructors to EpollEventLoopGroup which accept an Executor as a parameter
Result:
EpollEventLoopGroup is more consistent with NioEventLoopGroup
Fixes https://github.com/netty/netty/issues/5161
Motivation:
Before release 4.1.0.Final we should update all our dependencies.
Modifications:
Update dependencies.
Result:
Up-to-date dependencies used.
Motivation:
Some applications may use alternative methods of loading the epoll JNI symbols. We should support this use case.
Modifications:
Attempt to use a side effect free JNI method. If that fails, load the library.
Result:
Fixes#5122
Motivation:
We missed to correctly retrieve the localAddress() after we called Socket.connect(..) and so the user would always see an incorrect address when calling EpollSocketChannel.localAddress().
Modifications:
- Ensure we always retrieve the localAddress() after we called Socket.connect(...) as only after this we will be able to receive the correct address.
- Add unit test
Result:
Correct and consistent behaviour across different transports (NIO/OIO/EPOLL).
Motivation:
441aa4c575 conditionally set the readFlag based upon if maybeMoreDataToRead is set. It is possible that the read flag will not be set, and nothing will be read by executeEpollInReadyRunnable and no actual data will be read even though the user requested it.
Modifications:
- Always set the readFlag in doBeginRead
- Make it so only a single epollInReadyRunnable can execute for a channel at a time
Result:
Less chance of missing read events in EPOLL transport.
Motivation:
OIO/NIO use a volatile variable to track if a read is pending. EPOLL does not use a volatile an executes a Runnable on the event loop thread to set readPending to false. These mechansims should be consistent, and not using a volatile variable is preferable because the variable is written to frequently in the event loop thread.
OIO also does not set readPending to false before each fireChannelRead operation and may result in reading more data than the user desires.
Modifications:
- OIO/NIO should not use a volatile variable for readPending
- OIO should set readPending to false before each fireChannelRead
Result:
OIO/NIO/EPOLL are more consistent w.r.t. readPending and volatile variable operations are reduced
Fixes https://github.com/netty/netty/issues/5069
Motivation:
441aa4c575 introduced a bug in transport-native-epoll where readPending is set to false before a read is attempted, but this should happen before fireChannelRead is called. The NIO transport also only sets the readPending variable to false on the first read in the event loop. This means that if the user only calls read() on the first channelRead(..) the select loop will still listen for read events even if the user does not call read() on subsequent channelRead() or channelReadComplete() in the same event loop run. If the user only needs 2 channelRead() calls then by default they will may get 14 more channelRead() calls in the current event loop, and then 16 more when the event loop is woken up for a read event. This will also read data off the TCP stack and allow the peer to queue more data in the local RECV buffers.
Modifications:
- readPending should be set to false before each call to channelRead()
- make NIO readPending set to false consistent with EPOLL
Result:
NIO and EPOLL transport set readPending to false at correct times which don't read more data than intended by the user.
Fixes https://github.com/netty/netty/issues/5082
Motivation:
There is a spelling error in FileRegion.transfered() as it should be transferred().
Modifications:
Deprecate old method and add a new one.
Result:
Fix typo and can remove the old method later.
Motivation:
bfbef036a8 made EPOLL respect autoRead while in ET mode. However it is possible that we may miss data pending on the RECV queue if autoRead is off. This is because maybeMoreDataToRead is updated after fireChannelRead and if a user calls read() from here maybeMoreDataToRead will be false because it is updated after the fireChannelRead call. The way maybeMoreDataToRead was updated also causes a single channel to continuously read on the event loop and not relinquish and give other channels to try reading.
Modifications:
- Ensure maybeMoreDataToRead is always set after all user events, and is evaluated with readPending to execute a epollInReady on the EventLoop
- Combine the checkResetEpollIn and maybeMoreDataToRead logic to invoke a epollInReady later into the epollInFinally method due to similar responsibilities
- Update unit tests to reflect the user calling read() on the event loop from channelRead()
Result:
EPOLL ET with autoRead set to false will not leave data on the RECV queue.
Motivation:
Setting the WRITE_BUFFER_LOW_WATER_MARK before WRITE_BUFFER_HIGH_WATER_MARK results in an internal Exception (appears only in the logs) if the value is larger than the default high water mark value. The WRITE_BUFFER_HIGH_WATER_MARK call appears to have no effect in this context.
Setting the values in the reverse order works.
Modifications:
- deprecated ChannelOption.WRITE_BUFFER_HIGH_WATER_MARK and
ChannelOption.WRITE_BUFFER_LOW_WATER_MARK.
- add one new option called ChannelOption.WRITE_BUFFER_WATER_MARK.
Result:
The high/low water mark values limits caused by default values are removed.
Setting the WRITE_BUFFER_LOW_WATER_MARK before WRITE_BUFFER_HIGH_WATER_MARK results in an internal Exception (appears only in the logs) if the value is larger than the default high water mark value. The WRITE_BUFFER_HIGH_WATER_MARK call appears to have no effect in this context.
Setting the values in the reverse order works.
Motivation:
NIO now supports a pluggable select strategy, but EPOLL currently doesn't support this. We should strive for feature parity for EPOLL.
Modifications:
- Add SelectStrategy to EPOLL transport.
Result:
EPOLL transport supports SelectStategy.
Motivation:
We need to break out of the read loop for two reasons:
- If the input was shutdown in between (which may be the case when the user did it in the
fireChannelRead(...) method we should not try to read again to not produce any
miss-leading exceptions.
- If the user closes the channel we need to ensure we not try to read from it again as
the filedescriptor may be re-used already by the OS if the system is handling a lot of
concurrent connections and so needs a lot of filedescriptors. If not do this we risk
reading data from a filedescriptor that belongs to another socket then the socket that
was "wrapped" by this Channel implementation.
Modification:
Break the reading loop if the input was shutdown from within the channelRead(...) method.
Result:
No more meaningless exceptions and no risk to read data from wrong socket after the original was closed.
Motivation:
8dbf5d02e5 modified the shutdown code for Socket but did not correctly calculate the change in shutdown state and only applying this change. This is significant because if sockets are being opening and closed quickly and the underlying FD happens to be reused we need to take care that we don't unintentionally change the state of the new FD by acting on an object which represents the old incarnation of that FD.
Modifications:
- Calculate the shutdown change, and only apply what has changed, or exit if no change.
Result:
Socket.shutdown can not inadvertently affect the state of another logical FD.
Motivation:
cf171ff525 introduced a change in behavior when dealing with closing channel in the read loop. This changed behavior may use stale state to determine if a channel should be shutdown and may be incorrect.
Modifications:
- Revert the usage of potentially stale state
Result:
Closing a channel in the read loop is based upon current state instead of potentially stale state.
Motivation:
The code of transport-native-epoll missed some things in terms of static keywords, @deprecated annotations and other minor things.
Modifications:
- Add missing @deprecated annotation
- Not using FQCN in javadocs
- Add static keyword where possible
- Use final fields when possible
- Remove throws IOException from method where it is not needed.
Result:
Cleaner code.
Motivation:
In commit acbca192bd we changed to have our native operations which either gall getsockopt or setsockopt throw IOExceptions (to be more specific we throw a ClosedChannelException in some cases). Unfortunally I missed to also do the same for getSoError() and missed to add throws IOException to the native methods.
Modifications:
- Correctly throw IOException from getSoError()
- Add throws IOException to native methods where it was missed.
Result:
Correct declaration of getSoError() and other native methods.
Motivation:
If SO_LINGER is set to 0 the EPOLL transport will send a FIN followed by a RST. This is not consistent with the behavior of the NIO transport. This variation in behavior can cause protocol violations in streaming protocols (e.g. HTTP) where a FIN may be interpreted as a valid end to a data stream, but RST may be treated as the data is corrupted and should be discarded.
https://github.com/netty/netty/issues/4170 Claims the behavior of NIO always issues a shutdown when close occurs. I could not find any evidence of this in Netty's NIO transport nor in the JDK's SocketChannel.close() implementation.
Modifications:
- AbstractEpollChannel should be consistent with the NIO transport and not force a shutdown on every close
- FileDescriptor to keep state in a consistent manner with the JDK and not allow a shutdown after a close
- Unit tests for NIO and EPOLL to ensure consistent behavior
Result:
EPOLL is capable of sending just a RST to terminate a connection.
Motivation:
To be consistent with the JDK we should ensure our native methods throw a ClosedChannelException if the Channel was previously closed. This will then be wrapped in a ChannelException as usual. For all other errors we continue to just throw a ChannelException directly.
Modifications:
Ensure getsockopt and setsockopt will throw a ClosedChannelException if the channel was closed before, on other errors we throw a ChannelException as before diretly.
Result:
Consistent with the NIO Channel implementations.
Motivation:
We should always first notify the promise before trigger an event through the pipeline to be consistent.
Modifications:
Ensure we notify the promise before fire event.
Result:
Consistent behavior
Motivation:
EpollServerSocketConfig.isFreebind() throws an exception when called.
Modifications:
Use the correct getsockopt arguments.
Result:
No more exception when call EpollServerSocketConfig.isFreebind()
Motivation:
TCP_MD5 is only supported by SocketChannels so remove it from EpollServerChannelConfig which is generic.
Modifications:
Remove invalid code.
Result:
Remove invalid / dead code.
Motivation:
EPOLL does not support autoread when in ET mode.
Modifications:
- EpollRecvByteAllocatorHandle should not unconditionally force reading just because ET is enabled
- AbstractEpollChannel and all derived classes which implement epollInReady must support a variable which indicates
there may be more data to read. The variable will be used when read is called to simulate a EPOLL wakeup and call epollInReady if necessary. This will ensure that if we don't read until EAGAIN that we will try to read again and not rely on EPOLL to notify us.
Result:
EPOLL ET supports auto read.
Motivation:
When using the native transport have support for TCP_DEFER_ACCEPT or / and TCP_QUICKACK can be useful.
Modifications:
- Add support for TCP_DEFER_ACCEPT and TCP_QUICKACK
- Ad unit tests
Result:
TCP_DEFER_ACCEPT and TCP_QUICKACK are supported now.
Motivation:
For on tests we expected a ConnectTimeoutException but used the default timeout of 10 seconds. This slows down testing.
Modifications:
Use connect timeout of 1 second in unit test.
Result:
Faster execution of unit test.
Motivation:
JNI_OnUnload(...) does not return anything (has void in its signature) so we should not try to return something.
Modifications:
Remove return.
Result:
Fix incorrect but harmless code.
Motivation:
netty_epoll_native.c uses dladdr in attempt to get the name of the library that the code is running in. However the address passed to this funciton (JNI_OnLoad) may not be unique in the context of the application which loaded it. For example if another JNI library is loaded this address may first resolve to the other JNI library and cause the path name parsing to fail, which will cause the library to fail.
Modifications:
- Pass an addresses which is local to the current library to dladdr
Result:
EPOLL JNI library can be loaded in an environment where multiple JNI libraries are loaded.
Fixes https://github.com/netty/netty/issues/4840
Motivation:
Currently our epoll native transport requires sun.misc.Unsafe and so we need to take this into account for Epoll.isAvailable().
Modifications:
Take into account if sun.misc.Unsafe is present.
Result:
Only return true for Epoll.isAvailable() if sun.misc.Unsafe is present.
Motivation:
If Netty's class files are renamed and the type references are updated (shaded) the native libraries will not function. The native epoll module uses implicit JNI bindings which requires the fully qualified java type names to match the method signatures of the native methods. This means EPOLL cannot be used with a shaded Netty.
Modifications:
- Make the JNI method registration dynamic
- support a system property io.netty.packagePrefix which must be prepended to the name of the native library (to ensure the correct library is loaded) and all class names (to allow classes to be correctly referenced)
- remove system property io.netty.native.epoll.nettyPackagePrefix which was recently added and the code to support it was incomplete
Result:
transport-native-epoll can be used when Netty has been shaded.
Fixes https://github.com/netty/netty/issues/4800
Motivation:
As we now can easily build static linked versions of tcnative it makes sense to run our netty build against all of them.
This helps to ensure our code works with libressl, openssl and boringssl.
Modifications:
Allow to specify -Dtcnative.artifactId= and -Dtcnative.version=
Result:
Easy to run netty build against different tcnative flavors.
Motivation:
When a wildcard address is used to bind a socket and ipv4 and ipv6 are usable we should accept both (just like JDK IO/NIO does).
Modifications:
Detect wildcard address and if so use in6addr_any
Result:
Correctly accept ipv4 and ipv6
Motivation:
transport-native-epoll finds java classes from JNI using fully qualified class names. If a shaded version of Netty is used then these lookups will fail.
Modifications:
- Allow a prefix to be appended to Netty class names in JNI code.
Result:
JNI code can be used with shaded version of Netty.
Motivation:
We should also be able to compile the native transport on 32bit systems.
Modifications:
Add cast to intptr_t for pointers
Result:
It's possible now to also compile on 32bit.
Motivation:
transport-native-epoll has its pom.xml encoding attribute set to ISO-8859-15. Because
of this gradle, and other dependency management systems, can't correctly resolve this
library from wherever it happens to be published.
Modifications:
netty/transport-native-epoll/pom.xml had its xml encoding changed to UTF-9
Result:
Gradle, and other dependency management systems, will now be able to correctly resolve this module.
Motivation:
Linux uses different socket options to set the traffic class (DSCP) on IPv6
Modifications:
Also set IPV6_TCLASS for IPv6 sockets
Result:
TrafficClass will work on IPv4 and IPv6 correctly
Motivation:
If an user will close a Socket / FileDescriptor multiple times we should handle the extra close operations as NOOP.
Modifications:
Only do the actual closing one time
Result:
No exception if close is called multiple times.
Motivation:
We missed to define the actual c function for isKeepAlive(...) and so throw UnsatisfieldLinkError.
Modifications:
- Add function
- Add unit test for Socket class
Result:
Correctly work isKeepAlive(...) when using native transport
Motivation:
We need to remove all registered events for a Channel from the EventLoop before doing the actual close to ensure we not produce a cpu spin when the actual close operation is delayed or executed outside of the EventLoop.
Modifications:
Deregister for events for NIO and EPOLL socket implementations when SO_LINGER is used.
Result:
No more cpu spin.
Motivation:
We should retain the original hostname when connect to a remote peer so the user can still query the origin hostname if getHostString() is used.
Modifications:
Compute a InetSocketAddress from the original remote address and the one returned by the Os.
Result:
Same behavior when using epoll transport and nio transport.
Motivation:
Fix a race-condition when closing NioSocketChannel or EpollSocketChannel while try to detect if a close executor should be used and the underlying socket was already closed. This could lead to an exception that then leave the channel / in an invalid state and so could lead to side-effects like heavy CPU usage.
Modifications:
Catch possible socket exception while try to get the SO_LINGER options from the underlying socket.
Result:
No more race-condition when closing the channel is possible with bad side-effects.
Motivation:
AbstractEpollStreamChannel has a queue which collects splice events. Splice is assumed not to be the most common use case of this class and thus the splice queue could be initialized in a lazy fashion to save memory. This becomes more significant when the number of connections grows.
Modifications:
- AbstractEpollStreamChannel.spliceQueue will be initialized in a lazy fashion
Result:
Less memory consumption for most use cases
Motivation:
We should use OneTimeTask where possible to reduce object creation.
Modifications:
Replace Runnable with OneTimeTask
Result:
Less object creation
Motivation:
If we have a lot of writes going on we currently need to lookup the IovArray for each Channel that does writes. This can have quite some perf overhead. We should not need to do this and just store a reference of the IovArray on the EpollEventLoop itself.
Modifications:
- Remove IoArrayThreadLocal
- Store the IoArray in the EventLoop itself
Result:
Less FastThreadLocal lookups
Motivation:
If ChannelOption.ALLOW_HALF_CLOSURE is true and the shutdown input operation fails we should not propagate this exception, and instead consider this socket's read as half closed.
Modifications:
- AbstractEpollChannel.shutdownInput should not propagate exceptions when attempting to shutdown the input, but instead should just close the socket
Result:
Users expecting a ChannelInputShutdownEvent will get this event even if the socket is already shutdown, and the shutdown operation fails.
Motivation:
The EPOLL module was not completly respecting the half closed state. It may have missed events, or procssed events when it should not have due to checking isOpen instead of the appropriate shutdown state.
Modifications:
- use FileDescriptor's isShutdown* methods instead of isOpen to check for processing events.
Result:
Half closed code in EPOLL module is more correct.
Motivation:
transport-native-epoll is designed to be specific to Linux. However there is native code that can be extracted out and made to work on more Unix like distributions. There are a few steps to be completely decoupled but the first step is to extract out code that can run in a more general Unix environment from the Linux specific code base.
Modifications:
- Move all non-Linux specific stuff from Native.java into the io.netty.channel.unix package.
- io.netty.channel.unix.FileDescriptor will inherit all the native methods that are specific to file descriptors.
- io_netty_channel_epoll_Native.[c|h] will only have code that is specific to Linux.
Result:
Code is decoupled and design is streamlined in FileDescriptor.
Motivation:
Java_io_netty_channel_epoll_Native_getSoError incorrectly returns the value from the get socket option function.
Modifications:
- return the value from the result of the get socket option call
Result:
Java_io_netty_channel_epoll_Native_getSoError returns the correct value.
Motivation:
If a RDHUP and IN event occurred at the same time it is possible we may not read all pending data on the channel. We should ensure we read data before processing the RDHUP event.
Modifications:
- Process the RDHUP event before the IN event.
Result:
Data will not be dropped.
Fixes https://github.com/netty/netty/issues/4317
Motivation:
EPOLL attempts to support half closed socket, but fails to call shutdown to close the read portion of the file descriptor.
Motivation:
- If half closed is supported shutting down the input should call underlying Native.shutdown(...) to make sure the peer is notified of the half closed state.
Result:
EPOLL half closed is more correct.
Motivation:
We should fail the build on warnings in the JNI/c code.
Modifications:
- Add GCC flag to fail build on warnings.
- Fix warnings (which also fixed a bug when using splice with offsets).
Result:
Better code quality.
Motivation:
We should call shutdown(...) on the socket before closing the filedescriptor to ensure it is closed gracefully.
Modifications:
Call shutdown(...) before close.
Result:
Sockets are gracefully shutdown when using native transport.
Motivation:
On ubuntu, InetAddress.getLocalHost() will return 127.0.1.1 this causes some tests to fail.
NetUtil.LOCALHOST4 is more portable.
Modifications:
Made changes in EpollSocketTcpMd5Test to make test passing on ubuntu.
Result:
EpollSocketTcpMd5Test now also passes on ubuntu.
Motivation:
The latest netty-tcnative fixes a bug in determining the version of the runtime openssl lib. It also publishes an artificact with the classifier linux-<arch>-fedora for fedora-based systems.
Modifications:
Modified the build files to use the "-fedora" classifier when appropriate for tcnative. Care is taken, however, to not change the classifier for the native epoll transport.
Result:
Netty is updated the the new shiny netty-tcnative.
Motivation:
writeBytes(...) missed to set EPOLLOUT flag when not all bytes were written. This could lead to have the EpollEventLoop not try to flush the remaining bytes once the socket becomes writable again.
Modifications:
- Move setting EPOLLOUT flag logic to one point so we are sure we always do it.
- Move OP_WRITE flag logic to one point as well.
Result:
Correctly try to write pending data if socket becomes writable again.
Motivation:
TCP Fast Open allows data to be carried in the SYN and SYN-ACK packets and consumed by the receiving end during the initial connection handshake, and saves up to one full round-trip time (RTT) compared to the standard TCP, which requires a three-way handshake (3WHS) to complete before data can be exchanged. This commit enables support for TFO on server sockets.
Modifications:
Added new Integer Option TCP_FASTOPEN in EpollChannelOption.
Added getters/setters in EpollServerChannelConfig for TCP_FASTOPEN.
Added way to check if TCP_FASTOPEN is supported on server in Native.
Added setting on socket opt TCP_FASTOPEN if value is set on channel options in doBind in EpollServerSocketChannel.
Enhanced EpollSocketTestPermutation to contain a permutation for server socket containing fast open.
Result:
Users of native-epoll can set TCP_FASTOPEN on server sockets and thus leverage fast connect features of RFC7413 if client is capable of it.
Conflicts:
transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollChannelOption.java
Motivation:
There are protocols (BGP, SXP), which are typically deployed with TCP
MD5 authentication to protect sessions from being hijacked/torn down by
third parties. This facility is not available on most operating systems,
but is typically present on Linux.
Modifications:
- add a new EpollChannelOption, which is write-only
- teach Epoll(Server)SocketChannel to track which addresses have keys
associated
- teach Native how to set the MD5 signature keys for a socket
Result:
Users of the native-epoll transport can set MD5 signature keys and thus
leverage RFC-2385 protection on TCP connections.
Motivation:
See #4174.
Modifications:
Modify transport-native-epoll to allow setting TCP_USER_TIMEOUT.
Result:
Hanging connections that are written into will get timeouted.
Conflicts:
transport-native-epoll/src/main/java/io/netty/channel/epoll/EpollChannelOption.java
Motivation:
In NIO and OIO we throw a ChannelException if a ChannelConfig operation fails. We should do the same with epoll to be consistent.
Modifications:
Use ChannelException
Result:
Consistent behaviour across different transport implementations.
Motivation:
When try to get SO_LINGER from a fd that is closed an Exception is thrown. We should only try to get SO_LINGER if the fd is still open otherwise an Exception is thrown that can be ignored anyway.
Modifications:
First check if the fd is still open before try to obtain SO_LINGER setting when get the closeExecutor. This is also the same that we do in the NIO transport.
Result:
No more exception when calling unsafe.close() on a channel that has a closed file descriptor.
Motivation:
The method implementions for setSoLinger(...) and setTrafficClass(...) were swapped by mistake.
Modifications:
Use the correct implementation for setSoLinger(...) and setTrafficClass(...)
Result:
Correct behaviour when setSoLinger(...) and setTrafficClass(...) are used with the epoll transport.
Motivation:
Commit cf171ff525 changed the way read operations were done. This change introduced a feedback loop between fireException and epollInReady.
Modifications:
- All EPOLL*Channel* classes should not call fireException and also continue to read. Instead a read operation should be executed on the eventloop (if the channel's input is not closed, and other conditions are satisfied)
Result:
Exception processing and channelRead will not be in a feedback loop.
Fixes https://github.com/netty/netty/issues/4091
Motivation:
Because of java custom UTF encoding, it was previously impossible to use
nul-bytes in domain socket names, which is required for abstract domain
sockets.
Modifications:
- Pass the encoded string byte array to the native code
- Modify native code accordingly to work with nul-bytes in the the
array.
- Move the string encoding to UTF-8 in java code.
Result:
Unix domain socket addresses will work properly if they contain nul-
bytes. Address encoding for these addresses changes from UTF-8-like to
real UTF-8.
Motivation:
If is enabled and a channel is half closed it is possible for the EPOLL event loop to get into an infinite loop by continuously being woken up on the EPOLLRDHUP event.
Modifications:
- Ensure that the EPOLLRDHUP event is unregistered for to prevent infinite loop.
Result:
1 less infinite loop.
Motivation:
We not set any optimization flag when compile native transport
Modification:
Add -O3 to CFLAGS to have GCC do optimizations
Result:
Ship optimized native code
Motiviation:
The current read loops don't fascilitate reading a maximum amount of bytes. This capability is useful to have more fine grain control over how much data is injested.
Modifications:
- Add a setMaxBytesPerRead(int) and getMaxBytesPerRead() to ChannelConfig
- Add a setMaxBytesPerIndividualRead(int) and getMaxBytesPerIndividualRead to ChannelConfig
- Add methods to RecvByteBufAllocator so that a pluggable scheme can be used to control the behavior of the read loop.
- Modify read loop for all transport types to respect the new RecvByteBufAllocator API
Result:
The ability to control how many bytes are read for each read operation/loop, and a more extensible read loop.
Motivation:
IP_FREEBIND allows to bind to addresses without the address up yet or even the interface configured yet.
Modifications:
Add support for IP_FREEBIND.
Result:
It's now possible to use IP_FREEBIND when using the native epoll transport.
Motivation:
It would be useful to support the Java `Map` interface in our primitive maps.
Modifications:
Renamed current methods to "pXXX", where p is short for "primitive". Made the template for all primitive maps extend the appropriate Map interface.
Result:
Fixes#3970
Motivation:
We missed to register for EPOLLRDHUP events when construct the EpollSocketChannel from an existing FileDescriptor. This could cause to miss connection-resets.
Modifications:
Add Native.EPOLLRDHUP to the events we are interested in.
Result:
Connection-resets are detected correctly.
Motivation:
Some glibc/kernel versions will trigger an EPOLLERR event to notify
about failed connect and not an EPOLLOUT. Also EPOLLERR may be triggered
when a connection is broke.
Modification:
React on EPOLLERR like if an EPOLLOUT / EPOLLIN was received, this will work in
all cases as we handle errors in EPOLLOUT / EPOLLIN anyway.
Result:
Correctly detect errors.
Motivation:
The unit tests should not fail due to using a channel option which is not supported by the underlying kernel.
Modifications:
- Ignore RuntimeExceptions which are thrown by JNI code when setsockopt or getsockopt fails.
Result:
Unit tests pass if socket option is not supported by kernel.
Motiviation:
TCP_NOTSENT_LOWAT is only supported in linux kernel 3.12 or newer. The addition of this socket option prevents older kernels from building.
Modifications:
- Conditionally define TCP_NOTSENT_LOWAT if it is not defined
Result:
Kernels older than 3.12 can still compile the EPOLL module.
Motiviation:
Linux provides the TCP_NOTSENT_LOWAT socket option. This can be used to control how much unsent data is queued in the tcp kernel buffers. This can be important when application level protocols (SPDY, HTTP/2) have their own priority mechanism and don't want data queued in the kernel.
Modifications:
- The epoll module will have an additional socket option TCP_NOTSENT_LOWAT
- There will be JNI methods to control the underlying linux socket option mechanism
Result:
Linux EPOLL module exposes the TCP_NOTSENT_LOWAT socket option.
Motivation:
the JNI function ThrowNew won't release any allocated memory.
The method exceptionMessage is allocating a new string concatenating 2 constant strings
What is creating a small leak in case of these exceptions are happening.
Modifications:
Added new methods that will use exceptionMessage and free resources accordingly.
I am also removing the inline definition on these methods as they could be reused by
other added modules (e.g. libaio which should be coming soon)
Result:
No more leaks in case of failures.
Motivation:
Due a bug we not correctly handled connection refused errors and so failed the connect promise with the wrong exception.
Beside this we some times even triggered fireChannelActive() which is not correct.
Modifications:
- Add testcase
- correctly detect connect errors
Result:
Correct and consistent handling.
Motivation:
When trying to write more then Integer.MAX_VALUE / SSIZE_MAX via writev(...) the OS may return EINVAL depending on the kernel or the actual OS (bsd / osx always return EINVAL). This will trigger an IOException.
Modifications:
Never try to write more then Integer.MAX_VALUE / SSIZE_MAX when using writev.
Result:
No more IOException when write more data then Integer.MAX_VALUE / SSIZE_MAX via writev.
Motivation:
When EPOLLRDHUP is received we need to try to read at least one time to ensure
that we read all pending data from the socket. Otherwise we may loose data.
Modifications:
- Ensure we read all data from socket
- Ensure file descriptor is closed on doClose() even if doDeregister() throws an Exception.
- Only handle either EPOLLRDHUP or EPOLLIN as only one is needed to detect connection reset.
Result:
No more data loss on connection reset.
Motivation:
When using epoll_ctl we should respect the return value and do the right thing depending on it.
Modifications:
Adjust java and native code to respect epoll_ctl return values.
Result:
Correct and cleaner code.
Motivation:
Linux supports splice(...) to transfer data from one filedescriptor to another without
pass data through the user-space. This allows to write high-performant proxy code or to stream
stuff from the socket directly the the filesystem.
Modification:
Add AbstractEpollStreamChannel.spliceTo(...) method to support splice(...) system call
Result:
Splice is now supported when using the native linux transport.
Conflicts:
transport-native-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java
Motivation:
Because of a bug we missed to fail the connect future when doClose() is called. This can lead to a future which is never notified and so may lead to deadlocks in user-programs.
Modifications:
Correctly fail the connect future when doClose() is called and the connection was not established yet.
Result:
Connect future is always notified.
Motivation:
Each different *ChannelOption did extend ChannelOption in 4.0, which we changed in 4.1. This is a breaking change in terms of the API so we need to ensure we keep the old hierarchy.
Modifications:
- Let all *ChannelOption extend ChannelOption
- Add back constructor and mark it as @deprecated
Result:
No API breakage between 4.0 and 4.1
Motivation:
As we missed to correctly handle EPOLLRDHUP we produce an IOException which is unnessary. This leads
to have exceptionCaught(...) methods called.
Modifications:
When EPOLLRDHUP was received just close the socket and fail all pending writes.
Result:
Correctly handle of EPOLLRDHUP and so not miss-leading exceptions.
Motivation:
When an error happens during loading the native library it may try to generate a new RuntimeException before the RuntimeException is loaded.
Modifications:
- Load RuntimeException as first
Result:
No more segfaults possible
Motivation:
On a system where ipv4 and ipv6 are supported a user may want to use -Djava.net.preferIPv4Stack=true to restrict it to use ipv4 only.
This is currently ignored with the epoll transport.
Modifications:
Respect java.net.preferIPv4Stack system property.
Result:
-Djava.net.preferIPv4Stack=true will have the effect the user is looking for.
Motivation:
Due a a regression that was introduced by b898bdd we failed to set the localAddress if the connect did not success directly.
Modifications:
Correct set localAddress in doConnect(...)
Result:
Be able to get the localAddress in all cases.
Motivation:
During 6b941e9bdb I introduced a regression that could cause an IllegalStateException.
A non-proper fix was commited as part of #3443. This commit add a proper fix.
Modifications:
Remove FileDescriptor.INVALID and add FileDescriptor.isOpen() as replacement. Once FileDescriptor.close() is called isOpen() will return false.
Result:
No more IllegalStateException caused by a close channel.
Motivation:
EpollDragramChannel never calls fireChannelActive after connect() which is a bug.
Modifications:
Correctly call fireChannelActive if needed
Result:
Correct behaviour
Motivation:
Before struct's were passed per value and not pointer. This did enforce a memory copy which is not needed.
Modifications:
- Use "const struct....*" as replacement
Result:
No more unnecessary memory copies
Motivation:
When create address from filedescriptor we may use incorrect byte order and so end up with an incorrect InetAddress.
Modification:
Not manually shift bytes
Result:
Correct address in all cases.
Motivation:
Because of a regression sometimes accept could produce an IllegalArgumentException
Modifications:
Correctly respect offset when decode port and scope id.
Result:
No more IllegalArgumentException
Motivation:
This is a regression that was introduced as part of 6b941e9bdb. The regression could produce an "infinity" triggering of IllegalStateException if a channel goes inactive while process the events for it.
Modifications:
Correctly check if the channel is still active before trigger the callbacks.
Result:
No more IllegalStateException
Motivation:
There is a small race in the native transport where an accept(...) may success but a later try to obtain the remote address from the fd may fail is the fd is already closed.
Modifications:
Let accept(...) directly set the remote address.
Result:
No more race possible.
Motivation:
When epoll LT is used and autoRead == false when entering epollIn() we need to return without reading any data.
Modifications:
Correctly respect autoRead == false if using epoll LT.
Result:
Consistent and correct behaviour.
Motivation:
In the native transport we should throw a pre-instanced IOException on connection reset while reading.
Modifications:
Correctly throw pre-instanced IOException when ECONNRESET is received
Result:
Less overhead on connection reset
Motivation:
As we plan to have other native transports soon (like a kqueue transport) we should move unix classes/interfaces out of the epoll package so we
introduce other implementations without breaking stuff before the next stable release.
Modifications:
Create a new io.netty.channel.unix package and move stuff over there.
Result:
Possible to introduce other native impls beside epoll.
Motivation:
Sometimes it's useful to be able to create a Epoll*Channel from an existing file descriptor. This is especially helpful if you integrade some c/jni code.
Modifications:
- Add extra constructor to Epoll*Channel implementations that take a FileDescriptor as an argument
- Make Rename EpollFileDescriptor to NativeFileDescriptor and make it public
- Also ensure we obtain the correct remote/local address when create a Channel from a FileDescriptor
Result:
It's now possible to create a FileDescriptor and instance a Epoll*Channel via it.
Motivation:
If SO_LINGER is used shutdownOutput() and close() syscalls will block until either all data was send or until the timeout exceed. This is a problem when we try to execute them on the EventLoop as this means the EventLoop may be blocked and so can not process any other I/O.
Modifications:
- Add AbstractUnsafe.closeExecutor() which returns null by default and use this Executor for close if not null.
- Override the closeExecutor() in NioSocketChannel and EpollSocketChannel and return GlobalEventExecutor.INSTANCE if getSoLinger() > 0
- use closeExecutor() in shutdownInput(...) in NioSocketChannel and EpollSocketChannel
Result:
No more blocking of the EventLoop if SO_LINGER is used and shutdownOutput() or close() is called.
Motivation:
Some of the methods are frequently called and so should be inlined if possible.
Modifications:
Give the compiler a hint that we want to inline these methods.
Result:
Better performance if inlined.
Motivation:
Older linux kernels have problems handling a large value for epoll_wait(...) and so wait for ever.
Modifications:
Adjust timeout on the fly if a too big value is passed in.
Result:
Correctly works also on older kernels.
Motivation:
The writeSpinCount was ignored in the epoll transport and it just kept on trying writing. This could cause unnessary cpu spinning if a slow remote peer was reading the data very very slow.
Modification:
- Correctly take writeSpinCount into account when writing.
Result:
Less cpu spinning when writing to a slow remote peer.
Motivation:
Fix regression introduced by 585ce1593f, which missed to set EPOLLRDHUP for all stream channels.
Modifications:
Correctly set EPOLLRDHUP for all stream channels in the AbstractEpollStreamChannel constructor.
Result:
No more test failures in EpollDomain*Channel tests.
Motivation:
Before we used a long[] to store the ready events, this had a few problems and limitations:
- An extra loop was needed to translate between epoll_event and our long
- JNI may need to do extra memory copy if the JVM not supports pinning
- More branches
Modifications:
- Introduce a EpollEventArray which allows to directly write in a struct epoll_event* and pass it to epoll_wait.
Result:
Better speed when using native transport, as shown in the benchmark.
Before:
[xxx@xxx wrk]$ ./wrk -H 'Connection: keep-alive' -d 120 -c 256 -t 16 -s scripts/pipeline-many.lua http://xxx:8080/plaintext
Running 2m test @ http://xxx:8080/plaintext
16 threads and 256 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 14.56ms 8.64ms 117.15ms 80.58%
Req/Sec 286.17k 38.71k 421.48k 68.17%
546324329 requests in 2.00m, 73.78GB read
Requests/sec: 4553438.39
Transfer/sec: 629.66MB
After:
[xxx@xxx wrk]$ ./wrk -H 'Connection: keep-alive' -d 120 -c 256 -t 16 -s scripts/pipeline-many.lua http://xxx:8080/plaintext
Running 2m test @ http://xxx:8080/plaintext
16 threads and 256 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 14.12ms 8.69ms 100.40ms 83.08%
Req/Sec 294.79k 40.23k 472.70k 66.75%
555997226 requests in 2.00m, 75.08GB read
Requests/sec: 4634343.40
Transfer/sec: 640.85MB
Motivation:
Netty uses edge-triggered epoll by default for performance reasons. The downside here is that a messagesPerRead limit can not be enforced correctly, as we need to consume everything from the channel when notified.
Modification:
- Allow to switch epoll modes before channel is registered
- Some refactoring to share more code
Result:
It's now possible to switch epoll mode.
Motiviation:
When using domain sockets on linux it is supported to recv and send file descriptors. This can be used to pass around for example sockets.
Modifications:
- Add support for recv and send file descriptors when using EpollDomainSocketChannel.
- Allow to obtain the file descriptor for an Epoll*Channel so it can be send via domain sockets.
Result:
recv and send of file descriptors is supported now.
Motivation:
Using Unix Domain Sockets can be very useful when communication should take place on the same host and has less overhead then using loopback. We should support this with the native epoll transport.
Modifications:
- Add support for Unix Domain Sockets.
- Adjust testsuite to be able to reuse tests.
Result:
Unix Domain Sockets are now support when using native epoll transport.
Motivation:
At the moment the max number of events that can be handled per epoll wakup was set during construction.
Modifications:
- Automatically increase the max number of events to handle
Result:
Better performance when a lot of events need to be handled without adjusting the code.
