Motivation:
Add an option (through a SelectStrategy return code) to have the Netty event loop thread to do busy-wait on the epoll.
The reason for this change is to avoid the context switch cost that comes when the event loop thread is blocked on the epoll_wait() call.
On average, the context switch has a penalty of ~13usec.
This benefits both:
The latency when reading from a socket
Scheduling tasks to be executed on the event loop thread.
The tradeoff, when enabling this feature, is that the event loop thread will be using 100% cpu, even when inactive.
Modification:
Added SelectStrategy option to return BUSY_WAIT
Epoll loop will do a epoll_wait() with no timeout
Use pause instruction to hint to processor that we're in a busy loop
Result:
When enabled, minimizes impact of context switch in the critical path
Motivation:
When writing an HTTP/2 HEADERS with END_STREAM=1, the application expects
the stream to be closed afterward. However, the write can fail locally
due to HPACK encoder and similar. When that happens we need to make sure
to issue a RST_STREAM otherwise the stream can be closed locally but
orphaned remotely. The RST_STREAM is typically handled by
Http2ConnectionHandler.onStreamError, which will only send a RST_STREAM
if that stream still exists locally.
There are two possible flows for trailers, one handled immediately and
one going through the flow controller. Previously they behaved
differently, with the immedate code calling the error handler after
closing the stream. The immediate code also used a listener for calling
closeStreamLocal while the flow controlled code did so immediately after
the write.
The two code paths also differed in their VoidChannelPromise handling,
but both were broken. The immediate code path called unvoid() only if
END_STREAM=1, however it could always potentially add a listener via
notifyLifecycleManagerOnError(). And the flow controlled code path
unvoided incorrectly, changing the promise completion behavior. It also
passed the wrong promise to closeStreamLocal() in FlowControlledBase.
Modifications:
Move closeStreamLocal handling after calls to onError. This is the
primary change.
Now call closeStreamLocal immediately instead of when the future
completes. This is the more likely correct behavior as it matches that
of DATA frames.
Fix all the VoidChannelPromise handling.
Result:
Http2ConnectionHandler.onStreamError sees the same state as the remote
and issues a RST_STREAM, properly cleaning up the stream.
Motivation
The EpollChannelConfig (same for KQueues) and its subclasses repeatetly declare their own channel field which leads to a 3x repetition for each config instance. Given the fields are protected or package-private it's exposing the code code to "field hiding" bugs.
Modifications
Use the the existing protected channel field from the DefaultChannelConfig class and simply cast it when needed.
Result
Fixes#8331
Motivation:
Before when on server-side we just called the X509KeyManager methods when handshake() was called the first time which is not quite correct as we may not have received the full SSL hello / handshake and so could not extra for example the SNI hostname that was requested.
OpenSSL exposes the SSL_CTX_set_cert_cb function which allows to set a callback which is executed at the correct moment, so we should use it. This also allows us to support more methods of ExtendedSSLSession easily.
Modifications:
- Make use of new methods exposed by netty-tcnative since https://github.com/netty/netty-tcnative/pull/388 to ensure we select the key material at the correct time.
- Implement more methods of ExtendedOpenSslSession
- Add unit tests to ensure we are able to retrieve various things on server-side in the X509KeyManager and so verify it is called at the correct time.
- Simplify code by using new netty-tcnative methods.
Result:
More correct implementation for server-side usage and more complete implemented of ExtendedSSLSession.
Motivation:
a208f6dc7c added a testcase which uses hostname validation which may not be supported by OpenSSL depending on the version that is used. We should check first before we try to use it.
Modifications:
Add assumeTrue(...) check to ensure hostname validation is supported before trying to run the test.
Result:
No more test-failures on OpenSSL versions < 1.0.2.
Motivation:
When a X509TrustManager is used while configure the SslContext the JDK automatically does some extra checks during validation of provided certs by the remote peer. We should do the same when our native implementation is used.
Modification:
- Automatically wrap a X509TrustManager and so do the same validations as the JDK does.
- Add unit tests.
Result:
More consistent behaviour. Fixes https://github.com/netty/netty/issues/6664.
* Use AuthoritativeDnsServerCache for creating the new redirect stream.
Motivation:
At the moment if a user wants to provide custom sorting of the nameservers used for redirects it needs to be implemented in two places. This is more complicated as it needs to be.
Modifications:
- Just delegate to the AuthoritativeDnsServerCache always as we fill it before we call newRedirectDnsServerStream anyway.
Result:
Easier way for the user to implement custom sorting.
* Add cache for CNAME mappings resolved during lookup of DNS entries.
Motivation:
If the CNAMEd hostname is backed by load balancing component, typically the final A or AAAA DNS records have small TTL. However, the CNAME record itself is setup with longer TTL.
For example:
* x.netty.io could be CNAMEd to y.netty.io with TTL of 5 min
* A / AAAA records for y.netty.io has a TTL of 0.5 min
In current Netty implementation, original hostname is saved in resolved cached with the TTL of final A / AAAA records. When that cache entry expires, Netty recursive resolver sends at least two queries — 1st one to be resolved as CNAME record and the 2nd one to resolve the hostname in CNAME record.
If CNAME record was cached, only the 2nd query would be needed most of the time. 1st query would be needed less frequently.
Modifications:
Add a new CnameCache that will be used to cache CNAMEs and so may reduce queries.
Result:
Less queries needed when CNAME is used.
Motivation:
The first EA builds for Java12 are released so we should allow to run with these in our docker-compose setup.
Modifications:
Add docker-compose configs for Java12.
Result:
Be able to run easily with Java12 as well.
Motivation:
4d1458604a did fix some leaks in SniClientTest but missed the ones in SniClientJava8TestUtil.
Modifications:
Correctly release SslContext.
Result:
No more leaks in SNI tests.
Motivation:
We should add some command to be able to run all tests with leak detection enabled. This will then be used on the CI during PR builds.
Modifications:
Add new docker-compose config to run with leak-detection enabled.
Result:
Easy way to enable leak detection while running tests via docker.
Motivation:
We need to ensure the Cumulator always releases the input buffer if it can not take over the ownership of it as otherwise it may leak.
Modifications:
- Correctly ensure the buffer is always released.
- Add unit tests.
Result:
Ensure buffer is always released.
Motivation
Applications should not depend on internal packages with Java 9 and later. This cause a warning now, but will break in future versions of Java.
Modification
This change adds methods to UnixResolverDnsServerAddressStreamProvider (following after #6844) that parse /etc/resolv.conf for domain and search entries. Then DnsNameResolver does not need to rely on sun.net.dns.ResolverConfiguration to do this.
Result
Fixes#8318. Furthermore, at least in my testing with Java 11, this also makes multiple search entries work properly (previously I was only getting the first entry).
Motivation:
First EA releases of Java12 are out we should be able to compile with these and run tests.
Modifications:
Add maven profile for java12.
Result:
Be able to use Java12
Motivation:
In theory our estimation of the needed buffer could be off and so we need to ensure we grow it if there is no space left.
Modifications:
Ensure we grow the buffer if there is no space left in there but we still have data to deflate.
Result:
Correctly deflate data in all cases.
Motivation:
I noticed that we had some errors showing up in a test (which did not fail it tho) because we tried to full-fill the promise multiples times.
Modifications:
Use trySuccess(...) as we may produce multiple exceptions.
Result:
Less errors during test-run.
Motivation:
Java9 added getStatusResponses() to ExtendedSSLSession which we should correctly support when possible.
Modifications:
Implement the method correctly.
Result:
More complete and correct implementation.
Motivation:
If you encode a SOCKS5 message like new DefaultSocks5CommandResponse(FAILURE, DOMAIN, "", 0)
you correctly get a result of 05010003000000.
But if the bndAddr is null, for example like new DefaultSocks5CommandResponse(FAILURE, DOMAIN)
the encoded result is 0501000301000000 which means the domain name has a length of one and consists of a 0-byte.
Modification:
With this commit it is also correctly encoded as a string of 0 length.
Result:
Correctly encode empty SOCKS5 address
Motivation:
Conscrypt 1.3.0 was just released and adds support for TLSv1.3
Modifications:
Update to 1.3.0
Result:
Use latest conscrypt during build / test.
Motivation:
Get charset from Content-Type header even it contains multiple parameters.
Modification:
Extract charset value from the charset parameter if it is not last.
Result:
Fixes#8273
Motivation:
6ed7c6c75d added support for ExtendedOpenSslSession but we did not override getStatusResponses(). This lead to test failures on java9.
Modifications:
Implement ExtendedOpenSslSession.getStatusResponses() so it just returns an empty list.
Result:
Test pass again on Java9.
Motivation:
6ed7c6c75d added a test which blindly assumed we can use a KeyManagerFactory all the time. This is only true if have OpenSSL 1.0.2 or later, which may not be the case.
Modifications:
Only use KeyManagerFactory in test if the OpenSSL version does support it.
Result:
More robust tests.
Motivation:
When an ExtendedSSLSession is used its possible to do more strict checking of the keys during handshake. We should do this whenever possible.
Modification:
- Return an ExtendedSSLSession when using client-mode and Java7+
- Add unit test
- Simplify unit tests
Result:
More consistent behaviour.
Motivation:
We should not try to cast the Channel to a DatagramChannel as this will cause a ClassCastException.
Modifications:
- Do not cast
- rethrow from constructor if we detect the registration failed.
- Add unit test.
Result:
Propagate correct exception.
Motivation:
The Epoll transport checks to see if there are any scheduled tasks
before entering epoll_wait, and resets the timerfd just before.
This causes an extra syscall to timerfd_settime before doing any
actual work. When scheduled tasks aren't added frequently, or
tasks are added with later deadlines, this is unnecessary.
Modification:
Check the *deadline* of the peeked task in EpollEventLoop, rather
than the *delay*. If it hasn't changed since last time, don't
re-arm the timer
Result:
About 2us faster on gRPC RTT 50pct latency benchmarks.
Before (2 runs for 5 minutes, 1 minute of warmup):
```
50.0%ile Latency (in nanos): 64267
90.0%ile Latency (in nanos): 72851
95.0%ile Latency (in nanos): 78903
99.0%ile Latency (in nanos): 92327
99.9%ile Latency (in nanos): 119691
100.0%ile Latency (in nanos): 13347327
QPS: 14933
50.0%ile Latency (in nanos): 63907
90.0%ile Latency (in nanos): 73055
95.0%ile Latency (in nanos): 79443
99.0%ile Latency (in nanos): 93739
99.9%ile Latency (in nanos): 123583
100.0%ile Latency (in nanos): 14028287
QPS: 14936
```
After:
```
50.0%ile Latency (in nanos): 62123
90.0%ile Latency (in nanos): 70795
95.0%ile Latency (in nanos): 76895
99.0%ile Latency (in nanos): 90887
99.9%ile Latency (in nanos): 117819
100.0%ile Latency (in nanos): 14126591
QPS: 15387
50.0%ile Latency (in nanos): 61021
90.0%ile Latency (in nanos): 70311
95.0%ile Latency (in nanos): 76687
99.0%ile Latency (in nanos): 90887
99.9%ile Latency (in nanos): 119527
100.0%ile Latency (in nanos): 6351615
QPS: 15571
```
Motivation:
When using Epoll based transport, allow applications to configure SO_BUSY_POLL socket option:
SO_BUSY_POLL (since Linux 3.11)
Sets the approximate time in microseconds to busy poll on a
blocking receive when there is no data. Increasing this value
requires CAP_NET_ADMIN. The default for this option is con‐
trolled by the /proc/sys/net/core/busy_read file.
The value in the /proc/sys/net/core/busy_poll file determines
how long select(2) and poll(2) will busy poll when they oper‐
ate on sockets with SO_BUSY_POLL set and no events to report
are found.
In both cases, busy polling will only be done when the socket
last received data from a network device that supports this
option.
While busy polling may improve latency of some applications,
care must be taken when using it since this will increase both
CPU utilization and power usage.
Modification:
Added SO_BUSY_POLL socket option
Result:
Able to configure SO_BUSY_POLL from Netty
Motivation:
We should just directly init the refCnt to 1 and not use the AtomicIntegerFieldUpdater.
Modifications:
Just assing directly to 1.
Result:
Cleaner code and possible a bit faster as the JVM / JIT may be able to optimize the first store easily.
Motiviation:
At the moment whenever ensureAccessible() is called in our ByteBuf implementations (which is basically on each operation) we will do a volatile read. That per-se is not such a bad thing but the problem here is that it will also reduce the the optimizations that the compiler / jit can do. For example as these are volatile it can not eliminate multiple loads of it when inline the methods of ByteBuf which happens quite frequently because most of them a quite small and very hot. That is especially true for all the methods that act on primitives.
It gets even worse as people often call a lot of these after each other in the same method or even use method chaining here.
The idea of the change is basically just ue a non-volatile read for the ensureAccessible() check as its a best-effort implementation to detect acting on already released buffers anyway as even with a volatile read it could happen that the user will release it in another thread before we actual access the buffer after the reference check.
Modifications:
- Try to do a non-volatile read using sun.misc.Unsafe if we can use it.
- Add a benchmark
Result:
Big performance win when multiple ByteBuf methods are called from a method.
With the change:
UnsafeByteBufBenchmark.setGetLongUnsafeByteBuf thrpt 20 281395842,128 ± 5050792,296 ops/s
Before the change:
UnsafeByteBufBenchmark.setGetLongUnsafeByteBuf thrpt 20 217419832,801 ± 5080579,030 ops/s
* Log the correct line-number when using SLF4j with netty if possible.
Motivation:
At the moment we do not log the correct line number in many cases as it will log the line number of the logger wrapper itself. Slf4j does have an extra interface that can be used to filter out these nad make it more usable with logging wrappers.
Modifications:
Detect if the returned logger implements LocationAwareLogger and if so make use of its extra methods to be able to log the correct origin of the log request.
Result:
Better logging when using slf4j.
Motivation:
We should use the latest jmh version which also supports -prof dtraceasm on MacOS.
Modifications:
Update to latest jmh version.
Result:
Better benchmark / profiling support on MacOS.
* PemPrivateKey.toPem(...) should throw IllegalArgumentException when PrivateKey which does not support encoding is used.
Motivation:
At the moment when a PrivateKey is used that does not support encoding we throw a NPE when trying to convert the key. We should better throw an IllegalArgumentException with the details about what key we tried to encode.
Modifications:
- Check if PrivateKey.getEncoded() returns null and if so throw an IllegalArgumentException
- Add unit test.
Result:
Better handling of non-supported PrivateKey implementations.
Motivation:
Some of the flags we used are not supported anymore on more recent JDK versions. We should just remove all of them and only keep what we really need. This may also reflect better what people use in production.
Modifications:
Remove some flags when running the benchmarks.
Result:
Benchmarks also run with JDK11.
Motivation:
It is sometimes useful to be able to run benchmarks easily from the commandline and passs different arguments / options here. We should support this.
Modifications:
Add the benchmark-jar profile which allows to generate such an "uber-jar" that can be used directly to run benchmarks as documented at http://openjdk.java.net/projects/code-tools/jmh/.
Result:
More flexible way to run benchmarks.
Motivation:
In Java8 and earlier we used reflection to replace the used key set if not otherwise told. This does not work on Java9 and later without special flags as its not possible to call setAccessible(true) on the Field anymore.
Modifications:
- Use Unsafe to instrument the Selector with out special set when sun.misc.Unsafe is present and we are using Java9+.
Result:
NIO transport produce less GC on Java9 and later as well.
Motivation:
In Java8 and earlier we used reflection to detect if unaligned access is supported. This fails in Java9 and later as we would need to change the accessible level of the method.
Lucky enough we can use Unsafe directly to read the content of the static field here.
Modifications:
Add special handling for detecting if unaligned access is supported on Java9 and later which does not fail due jigsaw.
Result:
Better and more correct detection on Java9 and later.
Motivation:
We need to reset the offset to 0 when we fail lazy because of a too long frame.
Modifications:
- Reset offset
- Add testcase
Result:
Fixes https://github.com/netty/netty/issues/8256.
Motivation:
Optimizing the Epoll channel needs an objective measure of how fast
it is.
Modification:
Add a simple, closed loop, ping-pong benchmark.
Result:
Benchmark can be used to measure #7816
Initial numbers:
```
Result "io.netty.microbench.channel.epoll.EpollSocketChannelBenchmark.pingPong":
22614.403 ±(99.9%) 797.263 ops/s [Average]
(min, avg, max) = (21093.160, 22614.403, 24977.387), stdev = 918.130
CI (99.9%): [21817.140, 23411.666] (assumes normal distribution)
Benchmark Mode Cnt Score Error Units
EpollSocketChannelBenchmark.pingPong thrpt 20 22614.403 ± 797.263 ops/s
```
Motivation:
The JVM isn't always able to hoist out/reduce bounds checking (due to ref counting operations etc etc) hence making it configurable could improve performances for most CPU intensive use cases.
Modifications:
Each AbstractByteBuf bounds check has been tested against a new static final configuration property similar to checkAccessible ie io.netty.buffer.bytebuf.checkBounds.
Result:
Any user could disable ByteBuf bounds checking in order to get extra performances.
Motivation:
At the moment we will just assume the correct version of log4j2 is used when we find it on the classpath. This may lead to an AbstractMethodError at runtime. We should not use log4j2 if the version is not correct.
Modifications:
Check on class loading if we can use Log4J2 or not.
Result:
Fixes#8217.
Motivation:
We need to implement remove() by ourselves to make it work on Java7 as otherwise it will throw an AbstractMethodError. This is a followup of c1a335446d.
Modifications:
Just implemented remove()
Result:
Works on Java7 as well.
Motivation:
c1a335446d reimplemented remove(...) and contains(...) in a way which made it not work anymore when used by the Selector.
Modifications:
Partly revert changes in c1a335446d.
Result:
Works again as expected
Motivation:
5aaa16b24c introduced a testcase for specific ciphersuites and checked if these are supported by our native implementation before running it. Unfortunally this is not good enough as even on the JDK it may not be supported on various JDK versions (like Java7). Beside this the test leaked buffers.
Modifications:
- Correctly check if ciphersuite is supported on each SslProvider before trying to run test.
- Fix buffer leaks.
Result:
Testsuite pass again on Java7 and others when -Pleak is used.
Motivation:
Our SelectedSelectionKeySet does not correctly implement various methods which can be done without any performance overhead.
Modifications:
Implement iterator(), contains(...) and remove(...)
Result:
Related to https://github.com/netty/netty/issues/8242.
Motivation:
Log4J2Logger had some code-duplication with AbstractInternalLogger
Modifications:
Reuse AbstractInternaLogger.EXCEPTION_MESSAGE in Log4J2Logger and so remove some code-duplication
Result:
Less duplicated code.
Motivation:
We should use the latest netty-tcnative release which contains a fix to correctly support DH based ciphers when using openssl 1.1.x
Modifications:
Update to latest netty-tcnative which has the fix.
Result:
Correctly support DH ciphers in all cases. Fixes https://github.com/netty/netty/issues/8165.
