Commit Graph

8137 Commits

Author SHA1 Message Date
Norman Maurer
67be7c5b9f Log why it was not possible to use ByteBuffer.cleaner
Motivation:

We should log why we can not use ByteBuffer.cleaner and so maybe allow the user to fix it.

Modifications:

- Use Unsafe to access the field
- Log the exception when we can not use ByteBuffer.cleaner

Result:

Easier to debug why using cleaner is not possible.
2017-02-17 07:34:34 +01:00
Scott Mitchell
5de4029b43 Checkstyle fix from 56694eb 2017-02-16 22:28:56 -08:00
Scott Mitchell
56694ebc0f Cleanup from fbf0e5f4dd
Motivation:
ThreadLocalInsecureRandom still referenced ThreadLocalRandom directly, but shouldn't.

Modifications:
ThreadLocalInsecureRandom should reference PlatformDependent#threadLocalRandom() in comments

Result:
Less usage of internal.ThreadLocalRandom.
2017-02-16 15:56:23 -08:00
Norman Maurer
fbf0e5f4dd Prefer JDK ThreadLocalRandom implementation over ours.
Motivation:

We have our own ThreadLocalRandom implementation to support older JDKs . That said we should prefer the JDK provided when running on JDK >= 7

Modification:

Using ThreadLocalRandom implementation of the JDK when possible.

Result:

Make use of JDK implementations when possible.
2017-02-16 15:44:00 -08:00
Norman Maurer
38496a23da Update to netty-tcnative 2.0.0.Beta4 2017-02-16 15:42:19 -08:00
Scott Mitchell
795f318c3c Use a single array in SelectedSelectionKeySet
Motivation:
SelectedSelectionKeySet currently uses 2 arrays internally and users are expected to call flip() to access the underlying array and switch the active array. However we do not concurrently use 2 arrays at the same time and we can get away with using a single array if we are careful about when we reset the elements of the array.

Modifications:
- Introduce SelectedSelectionKeySetSelector which wraps a Selector and ensures we reset the underlying SelectedSelectionKeySet data structures before we select
- The loop bounds in NioEventLoop#processSelectedKeysOptimized can be defined more precisely because we know the real size of the underlying array

Result:
Fixes https://github.com/netty/netty/issues/6058
2017-02-16 15:10:37 -08:00
Scott Mitchell
4d7d478a3d Update JCTools to 2.0.1 2017-02-16 15:09:35 -08:00
Norman Maurer
6ac5f35077 Use Unsafe to read ByteBuffer.address field to make it work on Java9 as well.
Motivation:

Java9 does not allow changing access level via reflection by default. This lead to the situation that netty disabled Unsafe completely as ByteBuffer.address could not be read.

Modification:

Use Unsafe to read the address field as this works on all Java versions.

Result:

Again be able to use Unsafe optimisations when using Netty with Java9
2017-02-16 20:40:59 +01:00
Norman Maurer
c5324ea48c Make netty build work on Java9
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
2017-02-16 20:26:30 +01:00
Norman Maurer
ddd0e2f184 Enforce java8 for compilation
Motivation:

Commit 591293bfb4 changed the build to need java8 but missed to adjust the enforce rule as well.

Modifications:

Enforce java8+

Result:

Quickly fail when user tries to compile with pre java8
2017-02-16 20:24:54 +01:00
Scott Mitchell
0befcd8431 EpollDatagramChannel avoid getOption
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.
2017-02-16 11:04:37 -08:00
Scott Mitchell
f3dd410142 EPOLLRDHUP prematurely marking input as shutdown
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
2017-02-16 08:40:32 -08:00
Stephen E. Baker
9ee4cc0ada Correct comment for allowMaskMismatch parameter
Motivation:

The allowMaskMismatch parameter used throughout websocketx allows frames
with noncompliant masks when set to true, not false.

Modification:

Changed the javadoc comment everywhere it appears.

Result:

Fixes #6387
2017-02-16 17:03:55 +01:00
Norman Maurer
1843b31885 Guard against having NetworkInterface.getNetworkInterfaces() return null
Motivation:

NetworkInterface.getNetworkInterfaces() may return null if no network interfaces are found. We should guard against it.

Modifications:

Check for null return value.

Result:

Fixes [#6384]
2017-02-16 07:59:31 +01:00
Norman Maurer
8a3a3245df Ensure Unsafe buffer implementations are used when sun.misc.Unsafe is present
Motivation:

When sun.misc.Unsafe is present we want to use *Unsafe*ByteBuf implementations. We missed to do so in PooledByteBufAllocator when the heapArena is null.

Modifications:

- Correctly use UnpooledUnsafeHeapByteBuf
- Add unit tests

Result:

Use most optimal ByteBuf implementation.
2017-02-16 07:48:33 +01:00
Norman Maurer
b2f7e8648e Fix ReferenceCountedOpenSslEngine.getEnabledProtocols() when using boringssl
Motivation:

Commit cd3bf3df58 made netty observe the latest version of netty-tcnative which changed the way how static fields are computed for various SSL.* values. This lead to have SSL_OP_NO_SSLv2 become 0 when using boringssl as boringssl not supports SSLv2 at all. In the logic of ReferenceCountedOpenSslEngine.getEnabledProtocols() we not expect to have a zero value and so our logic fails.

Modifications:

Check we actual support the protocol before return it as enabled.

Result:

SSLEngineTest.testEnablingAnAlreadyDisabledSslProtocol passes again with boringssl
2017-02-16 07:47:06 +01:00
Scott Mitchell
4431ad894d OpenSslEngine may lose data if the non-application buffer is small/full
Motivation:
If an event occurs which generates non-application data (shutdown, handshake failure, alert generation, etc...) and the non-application buffer in the ByteBuffer BIO is full (or sufficiently small) we may not propagate all data to our peer before tearing down the socket.

Modifications:
- when wrap() detects the outbound is closed, but there is more data pending in the non-application buffers, we must also check if OpenSSL will generate more data from calling SSL_shutdown again
- when wrap() detects a handshakeExcpetion during failure we should check if OpenSSL has any pending data (in addition to the non-application buff) before throwing the handshake exception

Result:
OpenSslEngine more reliably transmits data to the peer before closing the socket.
2017-02-15 16:14:14 -08:00
Scott Mitchell
544d771152 Checkstyle cleanup from d59b4840c1 2017-02-15 12:03:11 -08:00
Jason Tedor
d59b4840c1 Cleanup ChannelId handling of basic methods
Motiviation:

Simplify implementation of compareTo/equals/hashCode for ChannelIds.

Modifications:

We simplfy the hashCode implementation for DefaultChannelId by not
making it random, but making it based on the underlying data. We fix the
compareTo implementation for DefaultChannelId by using lexicographic
comparison of the underlying data array. We fix the compareTo
implementation for CustomChannelId to avoid the possibility of overflow.

Result:

Cleaner code that is easier to maintain.
2017-02-15 11:53:36 -08:00
Norman Maurer
84188395be Remove backports of JDK8 classes
Motivation:

Java8 is out now for some time and JDK7 is no longer supported officially. We should remove all our backports and just use what the JDK provides us. This also will allow us to use intrinsics that are offered by the JDK implementations.

Modifications:

Remove all backports of jdk8 classes.

Result:

Use what the JDK offers us. This also fixes [#5458]
2017-02-15 20:44:23 +01:00
Norman Maurer
847359fd36 Fix incorrect delegate in overriden method in JdkOpenSslEngineInteroptTest
Motivation:

JdkOpenSslEngineInteroptTest.mySetupMutualAuthServerIsValidClientException(...) delegated to the wrong super method.

Modifications:

Fix delegate

Result:

Correct test-code.
2017-02-15 19:20:06 +01:00
Norman Maurer
43a2315372 Skip SSLEngineTests that depend on KeyManagerFactory when this is not supported by the openssl version.
Motivation:

Some version of openssl dont support the needed APIs to use a KeyManagerFactory. In this case we should skip the tests.

Modifications:

- Use assumeTrue(...) to skip tests that need a KeyManagerFactory and its not supported.

Result:

Tests pass on all openssl versions we support.
2017-02-15 19:17:43 +01:00
Jason Tedor
c92565d5c7 Correct expect header handling
Motivation:

Today, the HTTP codec in Netty responds to HTTP/1.1 requests containing
an "expect: 100-continue" header and a content-length that exceeds the
max content length for the server with a 417 status (Expectation
Failed). This is a violation of the HTTP specification. The purpose of
this commit is to address this situation by modifying the HTTP codec to
respond in this situation with a 413 status (Request Entity Too
Large). Additionally, the HTTP codec ignores expectations in the expect
header that are currently unsupported. This commit also addresses this
situation by responding with a 417 status.

Handling the expect header is tricky business as the specification (RFC
2616) is more complicated than it needs to be. The specification defines
the legitimate values for this header as "100-continue" and defines the
notion of expectatation extensions. Further, the specification defines a
417 status (Expectation Failed) and this is where implementations go
astray. The intent of the specification was for servers to respond with
417 status when they do not support the expectation in the expect
header.

The key sentence from the specification follows:

    The server MUST respond with a 417 (Expectation Failed) status if
    any of the expectations cannot be met or, if there are other
    problems with the request, some other 4xx status.

That is, a server should respond with a 417 status if and only if there
is an expectation that the server does not support (whether it be
100-continue, or another expectation extension), and should respond with
another 4xx status code if the expectation is supported but there is
something else wrong with the request.

Modifications:

This commit modifies the HTTP codec by changing the handling for the
expect header in the HTTP object aggregator. In particular, the codec
will now respond with 417 status if any expectation other than
100-continue is present in the expect header, the codec will respond
with 413 status if the 100-continue expectation is present in the expect
header and the content-length is larger than the max content length for
the aggregator, and otherwise the codec will respond with 100 status.

Result:

The HTTP codec can now be used to correctly reply to clients that send a
100-continue expectation with a content-length that is too large for the
server with a 413 status, and servers that use the HTTP codec will now
no longer ignore expectations that are not supported (any value other
than 100-continue).
2017-02-15 14:39:14 +01:00
Norman Maurer
b7acae03f2 Update tcnative package names
Motivation:

tcnative was moved into an internal package.

Modifications:

Update package for tcnative imports.

Result:

Use correct package names for tcnative.
2017-02-15 13:51:41 +01:00
Norman Maurer
5728e0eb2c Use the correct arguments when run with jdk9
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
2017-02-15 10:15:00 +01:00
Norman Maurer
c12908d608 Allow to compile and run tests with different java versions
Motivation:

As we now need to compile with java8 we should still allow to run the tests with a different java version to ensure everythin also works with java 7 and 6.

Modifications:

Allow to pass "-DtestJavaHome" to the build and so use a different java version during running the tests.

Result:

Be able to run tests with different java versions.
2017-02-15 10:11:43 +01:00
Scott Mitchell
d60e37cb3d OpenSslEngine wrap may not consume all data
Motivation:
If the OpenSslEngine has bytes pending in the non-application buffer and also generates wrapped data during the handshake then the handshake data will be missed. This will lead to a handshake stall and eventually timeout. This can occur if the non-application buffer becomes full due to a large certificate/hello message.

Modification:
- ReferenceCountedOpenSslEngine should not assume if no data is flushed from the non-application buffer that no data will be produced by the handshake.

Result:
New unit tests with larger certificate chains don't fail.
2017-02-15 09:29:32 +01:00
Scott Mitchell
c521c72178 SSLEngineTest cleanup
Motivation:
OpenSslEngineTest has unused imports and SSLEngineTest uses a fixed port which was used for debugging.

Modifications:
- Remove unused imports
- Use ephemeral port

Result:
Cleaner test code.
2017-02-14 17:28:13 -08:00
Scott Mitchell
84ebb4c315 Fix checkstyle issues introduced by fdcad3150e 2017-02-14 14:49:22 -08:00
Scott Mitchell
fdcad3150e Use tcnative's new setVerify modes
Modifications:
tcnative made some fixes and API changes related to setVerify. We should absorb these changes in Netty.

Modifications:
- Use tcnatives updated APIs
- Add unit tests to demonstrate correct behavior

Result:
Updated to latest tcnative code and more unit tests to verify expected behavior.
2017-02-14 12:14:58 -08:00
Scott Mitchell
cd3bf3df58 Consume tcnative options update
Motivation:
tcnative has updated how constants are defined and removed some constants which are either obsolete or now set directly in tcnative.

Modifications:
- update to compile against tcnative changes.

Result:
Netty compiles with tcnative options changes.
2017-02-14 12:09:10 -08:00
Norman Maurer
591293bfb4 Change minimum JDK version for compilation to 1.8
Motivation:

We previously changed netty to always compile with java7 as otherwise source compatibility was broken.

This was reported in [#3548] but was fixed in the meantime:
- https://bugs.openjdk.java.net/browse/JDK-8029240
- https://bugs.openjdk.java.net/browse/JDK-8030855

Modifications:

Change minimum JDK version for compilation to 1.8

Result:

Easier to maintain code.
2017-02-14 19:06:59 +01:00
Dmitriy Dumanskiy
506f0d8f8c Cleanup : String.length() == 0 replaced with String.isEmpty, removed unnecessary assert, class cast 2017-02-14 15:36:42 +01:00
Norman Maurer
90bc605477 Initialization of PlatformDependent0 fails on Java 9
Motivation:

Initialization of PlatformDependent0 fails on Java 9 in static initializer when calling setAccessible(true).

Modifications:

Add RefelectionUtil which can be used to safely try if setAccessible(true) can be used or not and if not fail back to non reflection.

Result:

Fixed [#6345]
2017-02-14 10:15:27 +01:00
Norman Maurer
78586a99b6 Ensure CombinedChannelDuplexHandler can not be shared.
Motivation:

CombinedChannelDuplexHandler must not be shared as it contains state.

Modifications:

Enforce that it is not shared.

Result:

Fixes [#6333]
2017-02-14 08:42:23 +01:00
Norman Maurer
fd2e142e74 Update to latest jmh version
Motivation:

We use an outdated jmh version.

Modifications:

Update to jmh 1.17.4.

Result:

Using latest jmh version.
2017-02-14 08:40:12 +01:00
Norman Maurer
f09a721d7f Expose the chunkSize used by PooledByteBufAllocator.
Motivation:

Sometimes it may be useful to know the used chunkSize.

Modifications:

Add method to expose chunkSize.

Result:

More exposed details.
2017-02-14 08:37:05 +01:00
Norman Maurer
371c0ca0f8 Eliminate unnessary wrapping when call ByteBuf.asReadOnly() in some cases
Motivation:

We can eliminate unnessary wrapping when call ByteBuf.asReadOnly() in some cases to reduce indirection.

Modifications:

- Check if asReadOnly() needs to create a new instance or not
- Add test cases

Result:

Less object creation / wrapping.
2017-02-14 08:35:16 +01:00
Norman Maurer
adcde84253 Allow to unwrap ByteBuffer > MAX_ENCRYPTED_PACKET_LENGTH
Motivation:

We should remove the restriction to only allow to call unwrap with a ByteBuffer[] whose cumulative length exceeds MAX_ENCRYPTED_PACKET_LENGTH.

Modifications:

Remove guard.

Result:

Fixes [#6335].
2017-02-14 08:30:35 +01:00
fenik17
0cf3f54a8d Adding 'final' keyword for private fields where possible
Motivation

Missing 'final' keyword for fields

Modifications

Add 'final' for fields where possible

Result

More safe and consistent code
2017-02-14 08:29:15 +01:00
Norman Maurer
974a251de8 Not fail tests when running on JDK9+ and init of MarshallingFactory fails
Motivation:

To use jboss-marshalling extra command-line arguments are needed on JDK9+ as it makes use of reflection internally.

Modifications:

Skip jboss-marshalling tests when running on JDK9+ and init of MarshallingFactory fails.

Result:

Be able to build on latest JDK9 release.
2017-02-14 08:27:58 +01:00
Norman Maurer
f7c8cf9cb9 Cleanup code in ssl package.
Motivation:

There were some warnings for the code in the ssl package.

Modifications:

- Remove not needed else blocks
- Use correctly base class for static usage
- Replace String.length() == 0 with String.isEmpty()
- Remove unused code

Result:

Less warnings and cleaner code.
2017-02-14 08:23:04 +01:00
Norman Maurer
34ea09e552 Add missing assumeTrue(...) that were not added in 54339c08ac 2017-02-14 08:17:33 +01:00
Norman Maurer
9b2b3e2512 Ensure tests pass when sun.misc.Unsafe is not present
Motivation:

We need to ensure we pass all tests when sun.misc.Unsafe is not present.

Modifications:

- Make *ByteBufAllocatorTest work whenever sun.misc.Unsafe is present or not
- Let Lz4FrameEncoderTest not depend on AbstractByteBufAllocator implementation details which take into account if sun.misc.Unsafe is present or not

Result:

Tests pass even without sun.misc.Unsafe.
2017-02-14 07:52:07 +01:00
Norman Maurer
54339c08ac Only try to calculate direct memory offset when sun.misc.Unsafe is present
Motivation:

We should only try to calculate the direct memory offset when sun.misc.Unsafe is present as otherwise it will fail with an NPE as PlatformDependent.directBufferAddress(...) will throw it.
This problem was introduced by 66b9be3a46.

Modifications:

Use offset of 0 if no sun.misc.Unsafe is present.

Result:

PooledByteBufAllocator also works again when no sun.misc.Unsafe is present.
2017-02-14 07:49:24 +01:00
Norman Maurer
d8596d2d90 Two tests are missing @Test annotations
Motivation:

ReadOnlyByteBufTest contains two tests which are missing the `@Test` annotation and so will never run.

Modifications:

Add missing annotation.

Result:

Tests run as expected.
2017-02-14 07:48:37 +01:00
Scott Mitchell
413d6eba53 EPOLL include error description and cause in exceptions
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.
2017-02-13 18:37:16 -08:00
Scott Mitchell
9ce74d46c1 Correct unit test flaw introduced in 54c9ecf682
Motivation:
54c9ecf682 introduced a unit tests which attempted to exclude addresses which resolved to loop back addresses from an assert statement. This was done with a static check for localhost but depending on machine configuration it is possible for other interfaces to be resolved.

Modifications:
- Use InetAddress#isLoopbackAddress() instead of string match on localhost

Result:
DnsNameResolverTest#testNameServerCache is more reliable.
2017-02-13 18:36:06 -08:00
Scott Mitchell
a1b5b5dcca EpollRecvByteAllocatorHandle doesn't inform delegate of more data
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.
2017-02-13 17:42:24 -08:00
Scott Mitchell
6765e9f99d CipherSuiteConverter NPE
Motivation:
CipherSuiteConverter may throw a NPE if a cipher suite from OpenSSL does not match the precomputed regular expression for OpenSSL ciphers. This method shouldn't throw and instead just return null.

Modifications:
- if cacheFromOpenSsl(..) fails the conversion toJava should return null

Result:
Fixes https://github.com/netty/netty/issues/6336.
2017-02-13 15:02:09 -08:00