Commit Graph

8853 Commits

Author SHA1 Message Date
Norman Maurer
4b728cd5bc
Ignore some test-flakiness when using Java11+ due outstanding Java11 bug. (#7984)
Motivation:

Java11 disallow draining any remaining bytes from the socket if a write causes a connection reset. This should be completely safe to do. At the moment if a write is causing a connection-reset you basically loose all the pending bytes that are sitting on the socket and are waiting to be read.

This happens because SocketOutputStream.write(…) may call AbstractPlainSocketImpl.setConnectionReset(…). Once this method is called any read(…) call will just throw a SocketException without even attempt to read any remaining data.

This is related:
 - https://bugs.openjdk.java.net/browse/JDK-8199329
 - http://hg.openjdk.java.net/jdk/jdk/rev/92cca24c8807
 - http://mail.openjdk.java.net/pipermail/net-dev/2018-May/011511.html

Modifications:

Tolarate if remaining bytes could not be read when using OIO.

Result:

Be able to build Netty and run testsuite while using Java11
2018-05-29 19:48:40 +02:00
Norman Maurer
0c6f077c18
Allow to test with boringssl-static and drop into shell (#7980)
Motivation:

We also need to run our tests while using boringssl-static to ensure everything works when using it. Beside this its sometimes useful to be able to just get a shell and so interactive work in the docker instance.

Modifications:

- Add configs for shell
- Add configs for testing with boringssl-static
- Ensure we not share .m2 when running tests

Result:

More complete docker setup.
2018-05-29 11:09:22 +02:00
Norman Maurer
2e587f75cb
Don't trim stacktrace for exceptions when running mvn test (#7981)
Motivation:

The maven surefire plugin will trim stacktraces by default which makes these kind of use-less when trying to understand why an test failed because one was thrown.

Modifications:

Configure the plugin to not trim the stacktrace.

Result:

Easier to debug test-failures.
2018-05-29 10:59:42 +02:00
Norman Maurer
f904c63a53
Correctly let Http2UnkownFrame extend HttpStreamFrame and so be usable with Http2MultiplexCodec. (#7976)
Motivation:

This is a followup for #7860. In the fix for #7860 we only partly fixed the problem as Http2UnknownFrame did not correctly extend HttpStreamFrame and so only worked when using the Http2FrameCodec. We need to have it extend HttpStreamFrame as otherwise Http2MultiplexCodec will reject to handle it correctly.

Modifications:

- Let Http2UnknownFrame extend HttpStreamFrame
- Add unit tests for writing and reading Http2UnkownFrame instances when the Http2MultiplexCodec is used.

Result:

Fixes https://github.com/netty/netty/issues/7969.
2018-05-29 07:27:40 +02:00
Norman Maurer
7f59896fba
Use jdk-11-ea+15 when try to build with java11 (#7979)
Motivation:

A new EA build for java 11 is out.

Modifications:

Update from ea+14 to ea+15

Result:

Use latest ea build
2018-05-28 11:51:25 +02:00
tomer doron
94946f6324 add docker-compose to allow easy testing across linux and java versions (#7958)
motivation: setup for testing across different permutations of linux and java versions

changes:
* refactor docker file to allow dynamic versions of centos and java
* add docker compose driver files for centos 6, 7 and java 1.8, 1.9, 1.10, 1.11
* update instructions
2018-05-28 10:31:48 +02:00
Nick Travers
9a3311506e Run the WebSocketClientHandshaker08Test on the 08 implementation (#7974)
Motivation:

The websockets abstract test suite does not run against the 08
implementation in the 08 version of the test suite.

Modifications:

Update the WebSocketClientHandshaker08Test to instantiate a new
WebSocketClientHandshaker08 rather than an 07 handshaker.

Result:

The WebSocketClientHandshaker08Test now tests the 08 implementation.
2018-05-27 10:02:49 +02:00
Norman Maurer
8a85761500
Don't use VM.maxDirectMemory() on IBM J9 / Eclipse OpenJ9 to retrieve direct memory limit (#7966)
Motivation:

On J9 / OpenJ9 netty initializes this value with 64M, even the direct accessible memory is actually unbounded.

Modifications:

Skip usage of VM.maxDirectMemory() on J9 / OpenJ9

Result:

More correct direct memory limit detection. Fixes #7654.
2018-05-25 14:35:32 +02:00
Norman Maurer
c3637ff42c
AUTO_CLOSE should not be marked as deprecated. (#7967)
Motivation:

A long time ago we deprecated AUTO_CLOSE but it turned out this feature is still useful because if a write error is detected there still maybe data to read, and if we close the channel automatically we will lose data

Modifications:

- Remove `@Deprecated` tag for AUTO_CLOSE, setAutoClose(...) and isAutoClose(...)
- Fix javadocs on ChannelConfig to correctly tell the default value of AUTO_CLOSE.

Result:

Less warnings.
2018-05-24 21:25:30 +02:00
Norman Maurer
030318e53c
Read until all data is consumed when EOF is detected even if readPend… (#7961)
* Read until all data is consumed when EOF is detected even if readPending is false and auto-read is disabled.

Motivation:

We should better always notify the user of EOF even if the user did not request any data as otherwise we may never be notified when the remote peer closes the connection. This should be ok as the amount of extra data we may read and so fire through the pipeline is limited by SO_RECVBUF.

Modifications:

- Always drain the socket when EOF is detected.
- Add testcase

Result:

No risk for the user to be not notified of EOF.
2018-05-24 20:29:29 +02:00
Nick Travers
19d1f4ea62 Propagate pong frames in WebSocketProtocolHandler (#7955)
Motivation:

Currently, on recipt of a PongWebSocketFrame, the
WebSocketProtocolHandler will drop the frame, rather than passing it
along so it can be referenced by other handlers.

Modifications:

Add boolean field to WebSocketProtocolHandler to indicate whether Pong
frames should be dropped or propagated, defaulting to "true" to preserve
existing functionality.

Add new constructors to the client and server implementations of
WebSocketProtocolHandler that allow for overriding the behavior for the
handling of Pong frames.

Result:

PongWebSocketFrames are passed along the channel, if specified.
2018-05-24 20:27:29 +02:00
Norman Maurer
4d6b006fe6
Correctly take status into account when compare DefaultHttpResponse (#7965)
Motivation:

DefaultHttpResponse did not respect its status when compute the hashCode and check for equality.

Modifications:

Correctly implement hashCode and equals

Result:

Fixes https://github.com/netty/netty/issues/7964.
2018-05-24 20:13:21 +02:00
Norman Maurer
583fc272f2
Fixed|SimpleChannelPool.close() should only return after complete. (#7927)
Motivation:

We need to ensure we only return from close() after all work is done as otherwise we may close the EventExecutor before we dispatched everything.

Modifications:

Correctly wait on operations to complete before return.

Result:

Fixes https://github.com/netty/netty/issues/7901.
2018-05-21 19:22:31 +02:00
Eric Anderson
88f0586a7e Remove HpackDecoder.maxHeaderListSizeGoAway (#7911)
Motivation:

When a sender sends too large of headers it should not unnecessarily
kill the connection, as killing the connection is a heavy-handed
solution while SETTINGS_MAX_HEADER_LIST_SIZE is advisory and may be
ignored.

The maxHeaderListSizeGoAway limit in HpackDecoder is unnecessary because
any headers causing the list to exceeding the max size can simply be
thrown away. In addition, DefaultHttp2FrameReader.HeadersBlockBuilder
limits the entire block to maxHeaderListSizeGoAway. Thus individual
literals are limited to maxHeaderListSizeGoAway.

(Technically, literals are limited to 1.6x maxHeaderListSizeGoAway,
since the canonical Huffman code has a maximum compression ratio of
.625. However, the "unnecessary" limit in HpackDecoder was also being
applied to compressed sizes.)

Modifications:

Remove maxHeaderListSizeGoAway checking in HpackDecoder and instead
eagerly throw away any headers causing the list to exceed
maxHeaderListSize.

Result:

Fewer large header cases will trigger connection-killing.
DefaultHttp2FrameReader.HeadersBlockBuilder will still kill the
connection when maxHeaderListSizeGoAway is exceeded, however.

Fixes #7887
2018-05-19 08:31:59 +02:00
Norman Maurer
987c443888
Use ByteBufAllocator used by the ReferenceCountedOpenSslEngine when build key-material. (#7952)
Motivation:

When we build the key-material we should use the ByteBufAllocator used by the ReferenceCountedOpenSslEngine when possible.

Modifications:

Whenever we have access to the ReferenceCountedOpenSslEngine we use its allocator.

Result:

Use correct allocator
2018-05-18 19:36:57 +02:00
Norman Maurer
7727649b2c
Add tests for the Conscrypt based SSLEngine. (#7950)
Motivation:

We currently have only interopt tests for Conscrypt, we should also have non-interopt tests.

Modifications:

Add ConscryptSslEngineTest

Result:

More tests
2018-05-18 19:36:40 +02:00
Norman Maurer
8ae126aaa8
Revert workaround in test for Java 11 as it produces a connection-reset as expected now. (#7951)
Motivation:

We added a workaround for Java 11 as it not produced a connect-reset when SO_LINGER with 0 was set and NIO was used. This was fixed in the latest ea release of Java 11:

 - http://hg.openjdk.java.net/jdk/jdk/rev/ea54197f4fe4
 - https://bugs.openjdk.java.net/browse/JDK-8203059

Modifications:

Revert workaround.

Result:

Test that Java 11 behave the same way as earlier Java versions again.
2018-05-18 12:27:56 +02:00
Norman Maurer
19e7b4438f
Update to conscrypt 1.1.2 (#7949)
Motivation:

We use latest conscrypt to test against.

Modifications:

Update to conscrypt 1.1.2

Result:

Use latest conscrypt release.
2018-05-18 12:27:40 +02:00
Norman Maurer
c3d29f7b9e
Guard against calling malloc(0) when create ByteBuffer. (#7948)
Motivation:

We did not guard against the case of calling malloc(0) when creating a ByteBuffer without a Cleaner. The problem is that malloc(0) can have different behaviour, it either return a null-pointer or a valid pointer that you can pass to free.

The real problem arise if Unsafe.allocateMemory(0) returns 0 and we use it as the memoryAddress of the ByteBuffer. The problem here is that native libraries test for 0 and handle it as a null-ptr. This is for example true in SSL.bioSetByteBuffer(...) which would throw a NPE when 0 is used as memoryAddress and so produced errors during SSL usage.

Modifications:

- Always allocate 1 byte as minimum (even if we ask for an empty buffer).
- Add unit test.

Result:

No more errors possible because of malloc(0).
2018-05-17 06:55:48 +02:00
Norman Maurer
546ddd2c28
Correctly calculate and respect if we can correctly fullfil wrap for alerts. (#7945)
Motivation:

We previously did not correctly take into account when we could not wrap (and so produce) the full SSL record with an alert when the SSLEngine was closed.

There are two problems here:

- If we call wrap(...) with an empty dst buffer after closeOutbound() was called we will not notify the user if we could not store the whole SSLRecord into the dst buffer and so we may produce incomplete SSLRecords

Modifications:

Add unit test which failed before.

Result:

Correctly handle the case when the dst buffer is not big enough and and alert needs to be produced.
2018-05-16 20:08:05 +02:00
Norman Maurer
47985c11c1 Add missing parameter when delegate to SSLEngine.
Motivation:

https://github.com/netty/netty/pull/7943 had a bug which caused to not have the argument passed to the delegating method.

Modifications:

Add argument to release call.

Result:

Correctly delegate method.
2018-05-16 20:01:58 +02:00
Norman Maurer
0bce0450c0
Adjust tests to also pass when using BoringSSL (#7946)
Motivation:

Some of the tests failed when using BoringSSL as some protocol / cipher combinations are not supported and it uses a different alert when the cert is not valid yet.

Modification:

- Remove protocol / cipher combos that are not supported by BoringSSL
- Test for different alert when using BoringSSL

Result:

Not test failures when using BoringSSL.
2018-05-16 18:58:27 +02:00
Norman Maurer
932d77b83e
Verify error stack is empty after each operation when using ReferenceCountedOpenSslEngine. (#7943)
Motivation:

https://github.com/netty/netty/pull/7941 proved that its easy to not correctly clear the error stack sometimes. We should do carefully test this.

Modifications:

Add a new SSLEngine wrapper that is used during tests, which verifies that the error stack is empty after each method call.

Result:

Better testing.
2018-05-16 13:50:37 +02:00
Norman Maurer
69c644bb98
Correctly detect if protocol is enabled when using netty-tcnative (#7940)
Motivation:

We sometimes did not correctly detect when a protocol is not enabled when using netty-tcnative as we did not take into account when the option flag is 0 (as for example in BoringSSL for SSLv2).

Modifications:

- Correctly take an option flag with 0 into account.
- Add unit tests.

Result:

Fixes https://github.com/netty/netty/issues/7935.
2018-05-16 07:23:55 +02:00
Norman Maurer
ed54ab034d
Correctly clear the error stack in all cases when using ReferenceCountedOpenSslEngine. (#7941)
Motivation:

We missed to correctly clear the error stack in one case when using the ReferenceCountedOpenSslEngine. Because of this it was possible to pick up an error on an unrelated operation.

Modifications:

- Correctly clear the stack
- Add verification of empty error stack to tests.

Result:

Not possible to observe unrelated errors.
2018-05-15 19:44:02 +02:00
Norman Maurer
1d09efeeb2
Correctly copy existing elements when CodecOutputList.add(index, element) is called. (#7939)
Motivation:

We did not correctly copy elements in some cases when add(index, element) was used.

Modifications:

- Correctly detect when copy is neede and when not.
- Add test case.

Result:

Fixes https://github.com/netty/netty/issues/7938.
2018-05-15 19:41:45 +02:00
ossdev07
85e2987719 pom.xml: adding suuport for AARCH64 architecture in Netty/Transport/N… (#7933)
Netty/Transport/Native/Epoll project can be build on aarch64 platform as well.

Motivation:
To provide the support for AARCH64 architecture

Modification:
Adjusted regex for enforce plugin to also allow AARCH64.

Result:
Be able to compile on AARCH64
2018-05-15 11:38:15 +02:00
Norman Maurer
3d2e231794 Add docker-sync files to .gitignore 2018-05-15 10:39:14 +02:00
Nick Hill
30a02b441a Avoid implicit allocations in Http2FrameLogger when logging is disabled (#7937)
Motivation:

Integer autoboxing in this class (and possibly also the varargs arrays)
showed non-negligible CPU and garbage contribution when profiling a gRPC
service. grpc-java currently hardcodes use of Http2FrameLogger, set at
DEBUG level.

Modifications:

Wrap offending log statements in conditional blocks.

Result:

Garbage won't be produced by Http2FrameLogger when set to a disabled
logging level.
2018-05-15 09:06:32 +02:00
zekaryu
09d9daf1c4 Update the comment of io.netty.buffer.PoolChunk.java (#7934)
Motivation:

When I read the source code, I found that the comment of PoolChunk is out of date, it may confuses readers with the description about memoryMap.

Modifications:

update the last passage of the comment of the PoolChunk class.

Result:
No change to any source code , just update comment.
2018-05-14 15:44:32 +02:00
Norman Maurer
64bb279f47 [maven-release-plugin] prepare for next development iteration 2018-05-14 11:11:45 +00:00
Norman Maurer
c67a3b0507 [maven-release-plugin] prepare release netty-4.1.25.Final 2018-05-14 11:11:24 +00:00
Norman Maurer
dc0cf3e099
Ensure Netty can be build with Java 11 (#7926)
Motivation:

Java 11 will be out soon, so we should be able to build (and run tests) netty.

Modifications:

- Add dependency that is needed till Java 11
- Adjust tests so these also pass on Java 11 (SocketChannelImpl.close() behavious a bit differently now).

Result:

Build also works (and tests pass) on Java 11.
2018-05-12 20:25:11 +02:00
Norman Maurer
c990c121ea
Only use jboss-marshalling when using Java 10 (#7929)
Motivation:

cff87de44c updated jboss-marshalling to 2.0.5.Final but this broke the ability to run tests with Java 7.

Modifications:

Only use 2.0.5.Final if compiled against Java 10 (as before 1.4.x works fine).

Result:

Be able to run tests with Java 7 on the CI.
2018-05-12 09:42:01 +02:00
Norman Maurer
cff87de44c
Make build pass on Java 10 (#7922)
Motivation:

Java 10 is out so we should be able to build netty with it (and run the tests).

Modifications:

- Update Mockito and JBoss Marshalling to support Java 10
- Fix unit test to not depend on specific cipher which is not present in Java 10 anymore

Result:

Netty builds (and runs all tests) when using Java 10
2018-05-10 08:36:17 +02:00
Karl Pauls
5256d17363 Make osgi testing work with java>=9 by updating to felix framework 5.6.10 (#7923)
Motivation:

Currently, the testing-osgi is set to skip if run with java>=9. That is not necessary when using a newer version of Felix.

Modification:

Update to Felix framework 5.6.10 (which has better jpms support), add some more --add-opens to not have WARN messages, and remove the skipOsgiTestsuite setting from the parent pom.

Result:

OSGi tests run and pass on java>=9.
2018-05-09 18:26:38 +02:00
Norman Maurer
369d64c3e6
Always follow cnames even if a matching A or AAAA record was found. (#7919)
Motivation:

At the moment if you do a resolveAll and at least one A / AAAA record is present we will not follow any CNAMEs that are also present. This is different to how the JDK behaves.

Modifications:

- Allows follow CNAMEs.
- Add unit test.

Result:

Fixes https://github.com/netty/netty/issues/7915.
2018-05-09 13:48:20 +02:00
Norman Maurer
cbe9ed8cc1
Add test that we handle thread.interrupt() in NioEventLoop (#7917)
Motivation:

We added some code to guard against thread.interrupt() in NioEventLoop but did not added a test.

Modifications:

Add testcase.

Result:

Verify that we correctly handle interrupt().
2018-05-09 08:57:53 +02:00
Norman Maurer
f01a590154
Enfore upper limit for minTtl when using DefaultCacheEntry. (#7920)
Motivation:

a598c3b69b added a upper limit for ttl but missed to also do the same for minTtl.

Modifications:

- Add upper limit for minTtl
- Add testcase.

Result:

No more IllegalArgumentException possible.
2018-05-09 08:57:20 +02:00
Xiaoyan Lin
a0ed6ec06c Fix the error message in ReferenceCounted.release (#7921)
Motivation:

When a buffer is over-released, the current error message of `IllegalReferenceCountException` is `refCnt: XXX, increment: XXX`, which is confusing. The correct message should be `refCnt: XXX, decrement: XXX`.

Modifications:

Pass `-decrement` to create `IllegalReferenceCountException`.

Result:

The error message will be `refCnt: XXX, decrement: XXX` when a buffer is over-released.
2018-05-08 20:09:16 +02:00
Konstantin Lutovich
437849bd93 Fix tiny typo in FixedChannelPool error message (#7918)
Motivation:

Closed `FixedChannelPool` fails acquire and release operations with
`IllegalStateException`s. These exceptions had message
"FixedChannelPooled was closed". Here "FixedChannelPooled" looks like
a typo and should probably be "FixedChannelPool".

Modifications:

Changed exception message to "FixedChannelPool was closed".

Result:

A tiny bit cleaner exception message.
2018-05-08 15:58:23 +02:00
Norman Maurer
18b170bd7a
Cleanup resolver-dns test code. (#7916)
Motivation:

We had some stuff that could be cleaned up in the resolver-dns test classes.

Modifications:

Cleanup code.

Result:

Cleaner code.
2018-05-08 08:55:17 +02:00
Norman Maurer
54c3de0f8a
Add test to validate SSLEngine does not zero out src buffer on wrap. (#7914)
Motivation:

We had a bug-report that claimed the src buffer used by OpenSslEngine will be zero out.

Modifications:

Add testcase to ensure correct behaviour

Result:

Testcase for https://github.com/netty/netty/issues/7753
2018-05-07 20:10:06 +02:00
Norman Maurer
d03bd44e9d
Add test for caching failed queries in DefaultDnsCache. (#7909)
Motivation:

We had no test that validated the handling of caching failures for DefaultDnsCache.

Modifications:

Add testcase.

Result:

More tests FTW.
2018-05-04 20:09:38 +02:00
Norman Maurer
358249e5c9
Allow to disable native transport and native ssl support via system property. (#7903)
Motivation:

Sometimes it's useful to disable native transports / native ssl to debug a problem. We should allow to do so with a system property so people not need to adjust code for this.

Modifications:

Add system properties which allow to disable native transport and native ssl.

Result:

Easier to disable native code usage without code changes.
2018-05-04 14:44:44 +02:00
Norman Maurer
b83328606f
DefaultDnsCache should store more then one Entry per hostname. (#7906)
Motivation:

Due a bug we did never store more then one address per hostname in DefaultDnsCache.

Modifications:

- Correctly store multiple entries per hostname
- Add tests

Result:

DefaultDnsCache correctly stores more then one entry. Also fixes https://github.com/netty/netty/issues/7882 .
2018-05-04 14:19:26 +02:00
Norman Maurer
a598c3b69b
Enforce sane upper limit for TTL in DefaultDnsCache. (#7907)
Motivation:

In b47fb81799 we limited the max supported delay to match what our internal implementat can support. Because of this it was possible that DefaultDnsCache produced an IllegalArgumentException when it tried to schedule a expiration > 3 years.

Modifications:

Limit the max supported TTL to 2 years which is safe for all our EventLoop implementations.

Result:

No more exceptions when adding records to the cache with a huge TTL.
2018-05-04 13:38:08 +02:00
Norman Maurer
095dadcbec
Fix race in ChannelReadHandler used during LocalChannel testing. (#7904)
Motivation:

ChannelReadHandler is used in tests added via f4d7e8de14. In the handler we verify the number of messages we receive per read() call but missed to sometimes reset the counter which resulted in exceptions.

Modifications:

Correctly reset read counter in all cases.

Result:

No more unexpected exceptions when running LocalChannel tests.
2018-05-04 07:36:33 +02:00
Norman Maurer
c60263e8a3
Correctly handle responses with status 205 and payload. (#7891)
Motivation:

HTTP responses with status of 205 should not contain a payload. We should enforce this.

Modifications:

Correctly handle responses with status 205 and payload by set Content-Length: 0 header and stripping out the content.

Result:

Fixes https://github.com/netty/netty/issues/7888
2018-05-03 11:25:06 +02:00
Daniel Schobel
bd800fa7e7 Add null-check to Htt2FrameCodec#consumeBytes. (#7899)
Motivation:

Streams can be deregistered so we can't assume their existence in the stream map.

Modifications:

Add a null-check in case a stream has been deregistered.

Result:

Fixes #7898.
2018-05-02 13:01:26 +02:00