Commit Graph

9462 Commits

Author SHA1 Message Date
Norman Maurer
ddc9f8bf1d Access the Constructor of the Channel in the constructor of ReflectiveChannelFactory. (#8718)
Motivation:

We should access the Constructor of the passed in class in the Constructor of ReflectiveChannelFactory only to reduce the overhead but also fail-fast.

Modifications:

Access the Constructor early.

Result:

Fails fast and less performance overhead.
2019-01-15 08:55:41 +01:00
Derek Lewis
f08e80ffd6 Remove unnecessary loop variable from AsciiString. (#8711)
Motivation:

Incrementing two variables in sync is not necessary when only one will do.

Modifications:

- Remove `j` from `for` loop and replace with `i`.
- Add more unit testing scenarios to cover changed code.

Results:

Unnecessary variable removed.
2019-01-15 08:33:48 +01:00
Norman Maurer
c10ccc5dec
Tighten contract between Channel and EventLoop by require the EventLoop on Channel construction. (#8587)
Motivation:

At the moment it’s possible to have a Channel in Netty that is not registered / assigned to an EventLoop until register(...) is called. This is suboptimal as if the Channel is not registered it is also not possible to do anything useful with a ChannelFuture that belongs to the Channel. We should think about if we should have the EventLoop as a constructor argument of a Channel and have the register / deregister method only have the effect of add a Channel to KQueue/Epoll/... It is also currently possible to deregister a Channel from one EventLoop and register it with another EventLoop. This operation defeats the threading model assumptions that are wide spread in Netty, and requires careful user level coordination to pull off without any concurrency issues. It is not a commonly used feature in practice, may be better handled by other means (e.g. client side load balancing), and therefore we propose removing this feature.

Modifications:

- Change all Channel implementations to require an EventLoop for construction ( + an EventLoopGroup for all ServerChannel implementations)
- Remove all register(...) methods from EventLoopGroup
- Add ChannelOutboundInvoker.register(...) which now basically means we want to register on the EventLoop for IO.
- Change ChannelUnsafe.register(...) to not take an EventLoop as parameter (as the EventLoop is supplied on custruction).
- Change ChannelFactory to take an EventLoop to create new Channels and introduce ServerChannelFactory which takes an EventLoop and one EventLoopGroup to create new ServerChannel instances.
- Add ServerChannel.childEventLoopGroup()
- Ensure all operations on the accepted Channel is done in the EventLoop of the Channel in ServerBootstrap
- Change unit tests for new behaviour

Result:

A Channel always has an EventLoop assigned which will never change during its life-time. This ensures we are always be able to call any operation on the Channel once constructed (unit the EventLoop is shutdown). This also simplifies the logic in DefaultChannelPipeline a lot as we can always call handlerAdded / handlerRemoved directly without the need to wait for register() to happen.

Also note that its still possible to deregister a Channel and register it again. It's just not possible anymore to move from one EventLoop to another (which was not really safe anyway).

Fixes https://github.com/netty/netty/issues/8513.
2019-01-14 20:11:13 +01:00
Derek Lewis
13f640cbba Updating ByteBuf Javadocs to represent actual behaviour. (#8709)
Motivation:

The javadocs stating `IndexOutOfBoundsException` is thrown were
different from what `ByteBuf` actually did. We want to ensure the
Javadocs represent reality.

Modifications:

Updated javadocs on `write*`, `ensureWriteable`, `capacity`, and
`maxCapacity` methods.

Results:

Javadocs more closely match actual behaviour.
2019-01-14 20:09:18 +01:00
Norman Maurer
9c725b91e2 Use OpenJDK 12 EA 27 when running CI jobs for JDK 12. (#8715)
Motivation:

A new EA release was done for OpenJDK12.

Modifications:

Use OpenJDK12 EA 27 when running CI jobs for JDK 12.

Result:

Test against latest OpenJDK 12 EA build.
2019-01-14 13:36:03 +01:00
Norman Maurer
4efad295df
Remove io.netty.channel.pool.* (#8681)
Motivation:

We should remove the ChannelPool and related implementations. It is often the case that having protocol knowledge can result in more effective pooling and ChannelPool currently doesn’t have this knowledge. This responsibility is assumed to be implemented at layers higher in the stack than Netty.

Modifications:

Remove io.netty.channel.pool.*

Result:

Less code to maintain, fixes https://github.com/netty/netty/issues/8549.
2019-01-14 08:28:40 +01:00
Norman Maurer
d06babf02a Correctly buffer multiple outbound streams if needed. (#8694)
Motivation:

In Http2FrameCodec we made the incorrect assumption that we can only have 1 buffered outboundstream as maximum. This is not correct and we need to account for multiple buffered streams.

Modifications:

- Use a map to allow buffer multiple streams
- Add unit test.

Result:

Fixes https://github.com/netty/netty/issues/8692.
2019-01-14 08:25:57 +01:00
Norman Maurer
d0891d08d7 Only call handlerRemoved(...) if handlerAdded(...) was called during adding the handler to the pipeline. (#8684)
Motivation:

Due a race in DefaultChannelPipeline / AbstractChannelHandlerContext it was possible to have only handlerRemoved(...) called during tearing down the pipeline, even when handlerAdded(...) was never called. We need to ensure we either call both of none to guarantee a proper lifecycle of the handler.

Modifications:

- Enforce handlerAdded(...) / handlerRemoved(...) semantics / ordering
- Add unit test.

Result:

Fixes https://github.com/netty/netty/issues/8676 / https://github.com/netty/netty/issues/6536 .
2019-01-14 08:20:02 +01:00
Norman Maurer
6ed296b719 Correctly detect and handle CNAME loops. (#8691)
Motivation:

We do not correctly detect loops when follow CNAMEs and so may try to follow it without any success.

Modifications:

- Correctly detect CNAME loops
- Do not cache CNAME entries which point to itself
- Add unit test.

Result:

Fixes https://github.com/netty/netty/issues/8687.
2019-01-14 08:17:51 +01:00
kashike
c0aa1ea5c7 Fix minor spelling issues in javadocs (#8701)
Motivation:

Javadocs contained some spelling errors, we should fix these.

Modification:

Fix spelling

Result:

Javadoc cleanup.
2019-01-14 07:25:13 +01:00
kezhenxu94
ef5b454a00 Use camel-case in NioEventLoop (#8713)
Motivation:

Java uses camel-case by convention.

Modification:

Consistently use camel-case.

Result:

More consistent code styling.
2019-01-14 07:22:41 +01:00
Jon Chambers
400b3081b0 Publicize default explicitFlushAfterFlushes count. (#8683)
Motivation:

Users who want to construct a `FlushConsolidationHandler` with a default `explicitFlushAfterFlushes` but non-default `consolidateWhenNoReadInProgress` may benefit from having an easy way to get the default "flush after flushes" count.

Modifications:

- Moved default `explicitFlushAfterFlushes` value to a public constant.
- Adjusted Javadoc accordingly.

Result:

Default `explicitFlushAfterFlushes` is accessible to callers.
2018-12-25 22:36:23 +01:00
Norman Maurer
71430d8bbd Call FastThreadLocal.removeAll() before notify termination future of … (#8666)
Motivation:

We should try removing all FastThreadLocals for the Thread before we notify the termination. future. The user may block on the future and once it unblocks the JVM may terminate and start unloading classes.

Modifications:

Remove all FastThreadLocals for the Thread before notify termination future.

Result:

Fixes https://github.com/netty/netty/issues/6596.
2018-12-21 11:11:05 +01:00
Alex Vasiliev
37e471dbe6 Added comments to LineBasedFrameDecoder, JsonObjectDecoder and XmlFrameDecoder that they are only compatible with UTF-8 encoded streams. (#8651)
Motivation:

LineBasedFrameDecoder, JsonObjectDecoder and XmlFrameDecoder upon investigation of the
sourcecode appeared to only support ASCII or UTF-8 input. It is an important characteristic
and ont reflected in any documentation. This could lead to improper usage and bugs.

Modifications:

Javadoc comment is addedd to all three classes to state that implementation is only
compatible with UTF-8 or ASCII input streams and brifly touches on implementaion details.

Result:

The end user of the netty library would not have to study sorcecode to deterime character
encoding limitations for given classes.
2018-12-20 07:40:29 +01:00
Bryce Anderson
44b919b698 Remove the maxChunkSize parameter for HTTP codecs (#8671)
Motivation:

The `maxChunkSize` only serves to split available content into even
smaller chunks which is not all that useful since the data is already
buffered.

Modification:

Remove the parameter.

Result:

Things are simpler.

Fixes #8430.
2018-12-20 07:37:51 +01:00
Norman Maurer
26e1411897 Add test for correctly handling SSLSessionBindingEvent when acting on th… (#8649)
Motivation:

During some other work I noticed we do not have any tests to ensure we correctly use SSLSessionBindingEvent. We should add some testing.

Modifications:

- Added unit test to verify we correctly implement it.
- Ignore the test when using Conscrypt as it not correctly implements it.

Result:

More tests for custom SSL impl.
2018-12-19 12:56:14 +01:00
Norman Maurer
5ecb34ee72 Fix ClassCastException and native crash when using kqueue transport. (#8665)
Motivation:

How we did the mapping from native code to AbstractKQueueChannel was not safe and could lead to heap corruption. This then sometimes produced ClassCastExceptions or could also lead to crashes. This happened sometimes when running the testsuite.

Modifications:

Use a Map for the mapping (just as we do in the native epoll transport).

Result:

No more heap corruption / crashes.
2018-12-19 12:14:11 +01:00
Norman Maurer
b12c3311ba Update to use OpenJDK 12 EA24 when building with Java 12 (#8672)
Motivation:

A new EA build was released for Java 12.

Modifications:

Update to OpenJDK 12 EA24

Result:

Use latest OpenJDK 12 build when building with Java 12
2018-12-19 11:32:04 +01:00
Stephane Landelle
9ed2884d3a Support 1012, 1013 and 1014 WebSocket close status code (#8664)
Motivation:

RFC 6455 doesn't define close status codes 1012, 1013 and 1014.
Yet, since then, IANA has defined them and web browsers support them.

From https://www.iana.org/assignments/websocket/websocket.xhtml:

* 1012: Service Restart
* 1013: Try Again Later
* 1014: The server was acting as a gateway or proxy and received an invalid response from the upstream server. This is similar to 502 HTTP Status Code.

Modification:

Make status codes 1012, 1013 and 1014 legit.

Result:

WebSocket status codes as defined by IANA are supported.
2018-12-17 19:43:10 +01:00
Norman Maurer
b32e9cea60 Upgrade to new version of autobahntestsuite maven plugin. (#8668)
Motivation:

A new version was released that fixes a few test-cases to allow more close codes.

Modifications:

Upgrade to 0.1.5

Result:

More compliant testing of websockets.
2018-12-17 17:26:00 +01:00
Norman Maurer
b1aefda2f6 Update to latest stable jython release (#8667)
Motivation:

Using the latest jython release fixes some noise that is produced by an exception that is thrown when jython is terminated.

Exception in thread "Jython-Netty-Client-4" Exception in thread "Jython-Netty-Client-7" Exception in thread "Jython-Netty-Client-5" java.lang.NoClassDefFoundError: org/python/netty/util/concurrent/DefaultPromise$2
        at org.python.netty.util.concurrent.DefaultPromise.notifyListeners(DefaultPromise.java:589)
        at org.python.netty.util.concurrent.DefaultPromise.setSuccess(DefaultPromise.java:397)
        at org.python.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:151)
        at java.lang.Thread.run(Thread.java:748)
Exception in thread "Jython-Netty-Client-8" java.lang.NoClassDefFoundError: org/python/netty/util/concurrent/DefaultPromise$2
        at org.python.netty.util.concurrent.DefaultPromise.notifyListeners(DefaultPromise.java:589)
        at org.python.netty.util.concurrent.DefaultPromise.setSuccess(DefaultPromise.java:397)
        at org.python.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:151)
        at java.lang.Thread.run(Thread.java:748)
Exception in thread "Jython-Netty-Client-3" java.lang.NoClassDefFoundError: org/python/netty/util/concurrent/DefaultPromise$2
        at org.python.netty.util.concurrent.DefaultPromise.notifyListeners(DefaultPromise.java:589)
        at org.python.netty.util.concurrent.DefaultPromise.setSuccess(DefaultPromise.java:397)%

Modification:

Update to latest stable release.

Result:

Less noise during build.
2018-12-17 10:25:18 +01:00
Norman Maurer
7f20adb5fc Skip tests that use KeyManagerFactory if not supported by OpenSSL version / flavor (#8662)
Motivation:

We missed to skip a few tests that depend on the KeyManagerFactory if the used OpenSSL version / flavor not support it.

Modifications:

Add missing overrides.

Result:

Testsuite also passes for example when using LibreSSL.
2018-12-14 21:34:05 +01:00
Norman Maurer
0efc47b69d NoClassDefFoundError on Android platform when try to use DefaultDnsServerAddressStreamProvider. (#8656)
Motivation:

Andoid does not contain javax.naming.* so we should not try to use it to prevent a NoClassDefFoundError on init.

Modifications:

Only try to use javax.naming.* to retrieve nameservers when not using Android.

Result:

Fixes https://github.com/netty/netty/issues/8654.
2018-12-14 21:32:57 +01:00
Norman Maurer
cb6ae72df2
Handling AUTO_READ should not be the responsibility of DefaultChannel… (#8650)
* Handling AUTO_READ should not be the responsibility of DefaultChannelPipeline but the Channel itself.

Motivation:

At the moment we do automatically call read() in the DefaultChannelPipeline when fireChannelReadComplete() / fireChannelActive() is called and the Channel is using auto read. This is nice in terms of sharing code but imho is not the responsibility of the ChannelPipeline implementation but the responsibility of the Channel implementation.

Modifications:

Move handing of auto read from DefaultChannelPipeline to Channel implementations.

Result:

More clear responsibiliy and not depending on implemention details of the ChannelPipeline.
2018-12-14 10:11:34 +00:00
Norman Maurer
05d481d8af Cleanup HTTP/2 tests for Http2FrameCodec and Http2MultiplexCodec (#8646)
Motiviation:

Http2FrameCodecTest and Http2MultiplexCodecTest were quite fragile and often not went through the whole pipeline which made testing sometimes hard and error-prone.

Modification:

- Refactor tests to have data flow through the whole pipeline and so made the test more robust (by testing the while implementation).

Result:

Easier to write tests for the codecs in the future and more robust testing in general.

Beside this it also fixes https://github.com/netty/netty/issues/6036.
2018-12-14 11:10:20 +01:00
Norman Maurer
a213fbba69 Fix typo in comment introduced by 83ab4ef5e3 2018-12-14 11:09:02 +01:00
Norman Maurer
75e6f597ce Explict always call ctx.read() when AUTO_READ is false and HTTP/2 is used. (#8647)
Motivation:

We should always call ctx.read() even when AUTO_READ is false as flow-control is enforced by the HTTP/2 protocol.

See also https://tools.ietf.org/html/rfc7540#section-5.2.2.

We already did this before but not explicit and only did so because of some implementation details of ByteToMessageDecoder. It's better to be explicit here to not risk of breakage later on.

Modifications:

- Ensure we always call ctx.read() when AUTO_READ is false
- Add unit test.

Result:

No risk of staling the connection when HTTP/2 is used.
2018-12-13 18:02:44 +00:00
Norman Maurer
7f22743d92
Revert "Ability to run a task at the end of an eventloop iteration." (#8637)
Motivation:

executeAfterEventLoopIteration is an Unstable API and isnt used in Netty. We should remove it to reduce complexity.

Changes:

This reverts commit 77770374fb.

Result:

Simplify implementation / cleanup.
2018-12-12 10:37:07 +01:00
Feri73
563793688f Adding support for whitespace in resource path in tests (#8606)
Motivation:

In windows if the project is in a path that contains whitespace,
resources cannot be accessed and tests fail.

Modifications:

Adds ResourcesUtil.java in netty-common. Tests use ResourcesUtil.java to access a resource.

Result:

Being able to build netty in a path containing whitespace
2018-12-12 10:29:19 +01:00
Norman Maurer
68f4c82d5a SSLSession.putValue / getValue / removeValue / getValueNames must be thread-safe. (#8648)
Motivation:

SSLSession.putValue / getValue / removeValue / getValueNames must be thread-safe as it may be called from multiple threads. This is also the case in the OpenJDK implementation.

Modifications:

Guard with synchronized (this) blocks to keep the memory overhead low as we do not expect to have these called frequently.

Result:

SSLSession implementation is thread-safe.
2018-12-12 07:41:31 +01:00
Norman Maurer
d9a6cf341c
Remove support for marking reader and writerIndex in ByteBuf to reduce overhead and complexity. (#8636)
Motivation:

ByteBuf supports “marker indexes”. The intended use case for these is if a speculative operation (e.g. decode) is in process the user can “mark” and interface and refer to it later if the operation isn’t successful (e.g. not enough data). However this is rarely used in practice,
requires extra memory to maintain, and introduces complexity in the state management for derived/pooled buffer initialization, resizing, and other operations which may modify reader/writer indexes.

Modifications:

Remove support for marking and adjust testcases / code.

Result:

Fixes https://github.com/netty/netty/issues/8535.
2018-12-11 14:00:49 +01:00
Paul Verest
9c594e5068 ReadTimeoutHandler - missing ) within JavaDoc example (#8645)
Motivation:

improve docs

Modification:

ReadTimeoutHandler - missing ) within JavaDoc example

No logic/unit tests affected
2018-12-10 20:50:43 +01:00
Norman Maurer
1881b8c781 Fix incorrect assert in Http2MultiplexCodec caused by 9f9aa1a.
Motivation:

9f9aa1a did some changes related to fixing how we handle ctx.read() in child channel but did incorrectly change some assert.

Modifications:

Fix assert to be correct.

Result:

Code does not throw an AssertionError due incorrect assert check.
2018-12-07 21:00:51 +01:00
Norman Maurer
0cc13cb0d4 Fix typo in MessageToMessageDecoder api docs. (#8638)
Motivation:

We had some typo (most likely caused by copy-and-paste) in the api docs which should be fixed.

Modifications:

Replace encoder by decoder word.

Result:

Correct apidocs.
2018-12-07 20:45:36 +01:00
Norman Maurer
10d7909013 More correct fix for using ChannelInitializer with custom EventExecutor. (#8633)
Motivation:

8331248671 did make some changes to fix a race in ChannelInitializer when using with a custom EventExecutor. Unfortunally these where a bit racy and so the testcase failed sometimes.

Modifications:

- More correct fix when using a custom EventExecutor
- Adjust the testcase to be more correct.

Result:

Proper fix for https://github.com/netty/netty/issues/8616.
2018-12-07 19:12:10 +01:00
多巴胺
0288570b06 Fix concurrency problem in UniqueIpFilter (#8635)
Motivation:

If two requests from the same IP are reached at the same time, `connected.contains(remoteIp)` may return false in both threads.

Modifications:

Check if there is already a connection with the same IP using return values.

Result:

Become thread safe.
2018-12-07 13:50:44 +01:00
tomer doron
4153f65c85 support publishing snapshots from docker based ci (#8634)
motivation: automate snapshot publishing from docker based ci

changes:
* add local settings.xml with env variables for publishing to sonatype-nexus-snapshots
* pipe UID/PWD env variable in docker compose
2018-12-07 06:16:39 +01:00
Norman Maurer
5e79d57b7d Skip test on windows as the semantics we expect are only true on Linux / Unix / BSD / MacOS (#8629)
Motivation:

In the test we assume some semantics on how RST is done that are not true for Windows so we should skip it.

Modifications:

Skip test when on windows.

Result:

Be able to run testsuite on windows. Fixes https://github.com/netty/netty/issues/8571.
2018-12-06 20:43:56 +01:00
Feri73
5618229203 Correcting Maven Dependencies (#8622)
Motivation:

Most of the maven modules do not explicitly declare their
dependencies and rely on transitivity, which is not always correct.

Modifications:

For all maven modules, add all of their dependencies to pom.xml

Result:

All of the (essentially non-transitive) depepdencies of the modules are explicitly declared in pom.xml
2018-12-06 09:02:00 +01:00
Norman Maurer
eb2118eeb4 ChannelInitializer may be invoked multiple times when used with custom EventExecutor. (#8620)
Motivation:

The ChannelInitializer may be invoked multipled times when used with a custom EventExecutor as removal operation may be done asynchronously. We need to guard against this.

Modifications:

- Change Map to Set which is more correct in terms of how we use it.
- Ensure we only modify the internal Set when the handler was removed yet
- Add unit test.

Result:

Fixes https://github.com/netty/netty/issues/8616.
2018-12-05 19:30:33 +01:00
Norman Maurer
564d5833cc NioEventLoop.register(...) should offload to the EventLoop if not alr… (#8612)
Motivation:

java.nio.channels.spi.AbstractSelectableChannel.register(...) need to obtain multiple locks during execution which may produce a long wait time if we currently select. This lead to multiple CI failures in the past.

Modifications:

Ensure the register call takes place on the EventLoop.

Result:

No more flacky CI test timeouts.
2018-12-05 15:31:49 +01:00
Norman Maurer
6a3723c619 Respect ctx.read() calls while processing reads for the child channels when using the Http2MultiplexCodec. (#8617)
Motivation:

We did not correct respect ctx.read() calls while processing a read for a child Channel. This could lead to read stales when auto read is disabled and no other read was requested.

Modifications:

- Keep track of extra read() calls while processing reads
- Add unit tests that verify that read() is respected when triggered either in channelRead(...) or channelReadComplete(...)

Result:

Fixes https://github.com/netty/netty/issues/8209.
2018-12-05 15:29:38 +01:00
Nick Travers
0a8c93a6fa Loosen bounds check on CompositeByteBuf's maxNumComponents (#8621)
Motivation:

In versions of Netty prior to 4.1.31.Final, a CompositeByteBuf could be
created with any size (including potentially nonsensical negative
values). This behavior changed in e7737b993, which introduced a bounds
check to only allow for a component size greater than one. This broke
some existing use cases that attempted to create a byte buf with a
single component.

Modifications:

Lower the bounds check on numComponents to include the single component
case, but still throw an exception for anything less than one.

Add unit tests for the case of numComponents being less than, equal to,
and greater than this lower bound.

Result:

Return to the behavior of 4.1.30.Final, allowing one component, but
still include an explicit check against a lower bound.

Note that while creating a CompositeByteBuf with a single component is
in some ways a contradiction of the term "composite", this patch caters
for existing uses while excluding the clearly nonsensical case of asking
for a CompositeByteBuf with zero or fewer components.

Fixes #8613.
2018-12-05 08:43:53 +01:00
Francesco Nigro
4c2b11633a Adding an execute burst cost benchmark for Netty executors (#8594)
Motivation:

Netty executors doesn't have yet any means to compare with each others
nor to compare with the j.u.c. executors

Modifications:

A new benchmark measuring execute burst cost is being added

Result:

It's now possible to compare some of Netty executors with each others
and with the j.u.c. executors
2018-12-04 15:46:48 +01:00
Norman Maurer
85c1590a90 Provide a way to cache the internal nioBuffer of the PooledByteBuffer… (#8603)
Motivation:

Often a temporary ByteBuffer is used which can be cached to reduce the GC pressure.

Modifications:

Cache the ByteBuffer in the PoolThreadCache as well.

Result:

Less GC.
2018-12-04 15:39:55 +01:00
Norman Maurer
361ace7671 Update to OpenJDK 12 ea22 (#8618)
Motivation:

We should use the latest OpenJDK 12 release when running tests against Java12.

Modifications:

- Update to OpenJDK 12 ea22.
- Update pax exam version
- skip OSGI testsuite on Java12 as it does not work ea22 yet.

Result:

Use latest OpenJDK 12 version when running on the CI.
2018-12-04 11:59:23 +01:00
Julien Hoarau
849bdc8a84 Set-Cookie headers should not be combined (#8611)
Motivation:

According to the HTTP spec set-cookie headers should not be combined
because they are not using the list syntax.

Modifications:

Do not combine set-cookie headers.

Result:

Set-Cookie headers won't be combined anymore
2018-12-01 10:47:37 +01:00
Nick Hill
0ecd0c6ff3 Reduce http2 buffer slicing (#8598)
Motivation

DefaultHttp2FrameReader currently does a fair amount of "intermediate"
slicing which can be avoided.

Modifications

Avoid slicing the input buffer in DefaultHttp2FrameReader until
necessary. In one instance this also means retainedSlice can be used
instead (which may also avoid allocating).

Results

Less allocations when using http2.
2018-11-29 19:46:19 +01:00
Nick Hill
fc0a4e15ab Harden ref-counting concurrency semantics (#8583)
Motivation

#8563 highlighted race conditions introduced by the prior optimistic
update optimization in 83a19d5650. These
were known at the time but considered acceptable given the perf
benefit in high contention scenarios.

This PR proposes a modified approach which provides roughly half the
gains but stronger concurrency semantics. Race conditions still exist
but their scope is narrowed to much less likely cases (releases
coinciding with retain overflow), and even in those
cases certain guarantees are still assured. Once release() returns true,
all subsequent release/retains are guaranteed to throw, and in
particular deallocate will be called at most once.

Modifications

- Use even numbers internally (including -ve) for live refcounts
- "Final" release changes to odd number (equivalent to refcount 0)
- Retain still uses faster getAndAdd, release uses CAS loop
- First CAS attempt uses non-volatile read
- Thread.yield() after a failed CAS provides a net gain

Result

More (though not completely) robust concurrency semantics for ref
counting; increased latency under high contention, but still roughly
twice as fast as the original logic. Bench results to follow
2018-11-29 08:32:50 +01:00
Norman Maurer
42e0ce6f7a Move less common code-path to extra method to allow inlining of writeUtf8. (#8600)
Motivation:

ByteBuf is used everywhere so we should try hard to be able to make things inlinable. During benchmarks it showed that writeCharSequence(...) fails to inline writeUtf8 because it is too big even if its hots.

Modifications:

Move less common code-path to extra method to allow inlining.

Result:

Be able to inline writeUtf8 in most cases.
2018-11-27 21:04:13 +01:00