Motivation:
The construction `new AsciiString(string.substring(...))` can be replaced with the `new AsciiString(string, start, length)` to avoid extra allocation.
Modifications:
Apply the described replacement in `HttpConversionUtil#setHttp2Authority`.
Result:
Less allocations.
Motivation:
Our http2 child channel implementation was not 100 % complete and had a few bugs. Beside this the performance overhead was non-trivial.
Modifications:
There are a lot of modifications, the most important....
* Http2FrameCodec extends Http2ConnectionHandler and Http2MultiplexCodec extends Http2FrameCodec to reduce performance heads and inter-dependencies on handlers in the pipeline
* Correctly handle outbound flow control for child channels
* Support unknow frame types in Http2FrameCodec and Http2MultiplexCodec
* Use a consistent way how to create Http2ConnectionHandler, Http2FrameCodec and Http2MultiplexCodec (via a builder)
* Remove Http2Codec and Http2CodecBuilder as the user should just use Http2MultipleCodec and Http2MultiplexCodecBuilder now
* Smart handling of flushes from child channels to reduce overhead
* Reduce object allocations
* child channels always use the same EventLoop as the parent Channel to reduce overhead and simplify implementation.
* Not extend AbstractChannel for the child channel implementation to reduce overhead in terms of performance and memory usage
* Remove Http2FrameStream.managedState(...) as the user of the child channel api should just use Channel.attr(...)
Result:
Http2MultiplexCodec (and so child channels) and Http2FrameCodec are more correct, faster and more feature complete.
Motivation:
This PR (unfortunately) does 4 things:
1) Add outbound flow control to the Http2MultiplexCodec:
The HTTP/2 child channel API should interact with HTTP/2 outbound/remote flow control. That is,
if a H2 stream used up all its flow control window, the corresponding child channel should be
marked unwritable and a writability-changed event should be fired. Similarly, a unwritable
child channel should be marked writable and a writability-event should be fired, once a
WINDOW_UPDATE frame has been received. The changes are (mostly) contained in ChannelOutboundBuffer,
AbstractHttp2StreamChannel and Http2MultiplexCodec.
2) Introduce a Http2Stream2 object, that is used instead of stream identifiers on stream frames. A
Http2Stream2 object allows an application to attach state to it, and so a application handler
no longer needs to maintain stream state (i.e. in a map(id -> state)) himself.
3) Remove stream state events, which are no longer necessary due to the introduction of Http2Stream2.
Also those stream state events have been found hard and complex to work with, when porting gRPC
to the Http2FrameCodec.
4) Add support for HTTP/2 frames that have not yet been implemented, like PING and SETTINGS. Also add
a Http2FrameCodecBuilder that exposes options from the Http2ConnectionHandler API that couldn't else
be used with the frame codec, like buffering outbound streams, window update ratio, frame logger, etc.
Modifications:
1) A child channel's writability and a H2 stream's outbound flow control window interact, as described
in the motivation. A channel handler is free to ignore the channel's writability, in which case the
parent channel is reponsible for buffering writes until a WINDOW_UPDATE is received.
The connection-level flow control window is ignored for now. That is, a child channel's writability
is only affected by the stream-level flow control window. So a child channel could be marked writable,
even though the connection-level flow control window is zero.
2) Modify Http2StreamFrame and the Http2FrameCodec to take a Http2Stream2 object intstead of a primitive
integer. Introduce a special Http2ChannelDuplexHandler that has newStream() and forEachActiveStream()
methods. It's recommended for a user to extend from this handler, to use those advanced features.
3) As explained in the documentation, a new inbound stream active can be detected by checking if the
Http2Stream2.managedState() of a Http2HeadersFrame is null. An outbound stream active can be detected
by adding a listener to the ChannelPromise of the write of the first Http2HeadersFrame. A stream
closed event can be listened to by adding a listener to the Http2Stream2.closeFuture().
4) Add a simple Http2FrameCodecBuilder and implement the missing frame types.
Result:
1) The Http2MultiplexCodec supports outbound flow control.
2) The Http2FrameCodec API makes it easy for a user to manage custom stream specific state and to create
new outbound streams.
3) The Http2FrameCodec API is much cleaner and easier to work with. Hacks like the ChannelCarryingHeadersFrame
are no longer necessary.
4) The Http2FrameCodec now also supports PING and SETTINGS frames. The Http2FrameCodecBuilder allows the Http2FrameCodec
to use some of the rich features of the Http2ConnectionHandler API.
Motivation:
When we use pipeline.replace and we still had ongoing inbound, then
there will be some problem that inbound message would go to wrong
handlers. So we add handler first, and remove self after add, so that
the next handler will be the correct one.
Modifications:
Uses remove after addAfter instead of replace.
Result:
Fixed#6881
Motivation:
There should be a way to allow graceful shutdown to wait for all open streams to close without a timeout. Using gracefulShutdownTimeoutMillis with a large value is a bit of a hack, and has a gotcha that sufficiently large values will overflow the long, resulting in a ClosingChannelFutureListener that executes immediately.
Modification:
Allow to use gracefulShutdownTimeoutMillis(-1) to express waiting until all streams are closed.
Result:
We can now shutdown the connection without a forced timeout.
Motivation:
codec-http2 currently does not strictly enforce the HTTP/1.x semantics with respect to the number of headers defined in RFC 7540 Section 8.1 [1]. We currently don't validate the number of headers nor do we validate that the trailing headers should indicate EOS.
[1] https://tools.ietf.org/html/rfc7540#section-8.1
Modifications:
- DefaultHttp2ConnectionDecoder should only allow decoding of a single headers and a single trailers
- DefaultHttp2ConnectionEncoder should only allow encoding of a single headers and optionally a single trailers
Result:
Constraints of RFC 7540 restricting the number of headers/trailers is enforced.
Motivation:
DefaultHttp2ConnectionEncoder#writeHeaders attempts to find a stream object, and if one doesn't exist it tries to create one. However in the event that the local endpoint has received a RST_STREAM frame before writing the response headers we attempt to create a stream. Since this stream ID is for the incorrect endpoint we then generate a GO_AWAY for what appears to be a protocol error, but can instead be failed locally.
Modifications:
- Just fail the local promise in the above situation instead of sending a GO_AWAY
Result:
Less severe consequences if the server asynchronously sends headers after a RST_STREAM has been received.
Fixes https://github.com/netty/netty/issues/6906.
Motivation:
PR #6811 introduced a public utility methods to decode hex dump and its parts, but they are not visible from netty-common.
Modifications:
1. Move the `decodeHexByte`, `decodeHexDump` and `decodeHexNibble` methods into `StringUtils`.
2. Apply these methods where applicable.
3. Remove similar methods from other locations (e.g. `HpackHex` test class).
Result:
Less code duplication.
Motivation:
Some JUnit assert calls can be replaced by simpler.
Modifications:
Replacement with a more suitable methods.
Result:
More informative JUnit reports.
Motivation:
DefaultHttp2FrameWriter has constructors that it would be a hassle to
expose as configuration parameters on Http2Codec. We should instead
make a builder for Http2Codec.
Modifications:
Get rid of the public constructors on Http2Codec and instead make sure
you can always use the builder where you would have used the constructor
before.
Result:
Http2Codec can be configured more flexibly, and the SensitivityDetector
can be configured.
Motivation:
Each StreamByteDistributor may allow for priority in different ways, but there are certain characteristics which are invalid regardless of the distribution algorithm. We should validate these invalid characteristics at the flow controller level.
Modifications:
- Disallow negative stream IDs from being used. These streams may be accepted by the WeightedFairQueueByteDistributor and cause state for other valid streams to be evicted.
- Improve unit tests to verify limits are enforced.
Result:
Boundary conditions related to the priority parameters are validated more strictly.
Motivation:
Chrome was randomly getting stuck loading the tiles examples.
Investigation showed that the Netty flow controller thought it had
nothing to send for the connection even though some streams has queued
data and window available.
Modifications:
Fixed an accounting error where an implicitly created parent was not
being added to the dependency tree, thus it and all of its children were
orphaned from the connection's tree and would never have data written.
Result:
Fixes#6621
Motivation:
The CI servers have reported leaks while building the HTTP/2 unit tests. The unit tests attempt to wait for the channels to be closed before exiting the test, but we should wait in case there are any tasks pending on the EventLoopGroup's task queues.
Modifications:
- Change the Future.sync() operations to Future.syncUninterruptibly()
- HTTP/2 unit tests which use local channel should wait for 5 seconds before shutting down the EventLoopGroups
Result:
More likely that any cleanup related tasks will execute before the unit tests are shutdown.
Motivation:
codec-http2 has some helper methods to write to ByteBuf in a big endian fashion. This is the default memory structure for ByteBuf so these helper methods are not necessary.
Modifications:
- remove writeUnsignedInt and writeUnsignedShort
Result:
codec-http2 has less ByteBuf helper methods which are not necessary.
Motivation:
HTTP/2 support two ways to start on a no-tls tcp connection,
http/1.1 upgrade and prior knowlege methodology to start HTTP/2.
Currently, the http2-server from example only support
starting by upgrade. I think we can do a simple dispatch by peek first
bytes from inbound that match to prior knowledge preface or not and
determine which handlers to set into pipeline.
Modifications:
Add ClearTextHttp2ServerUpgradeHandler to support start HTTP/2 via clear
text with two approach. And update example/http2-server to support
this functionality.
Result:
netty HTTP/2 and the example http2-server accept for two ways to start
HTTP/2 over clear text.
Fixed memory leak problem
Update fields to final
Rename ClearText to cleartext
Addressed comments for code improvement
- Always prefer static, final, and private if possible
- Add UnstableApi annotation
- Used EmbeddedChannel.readInbound instead of unhandled inbound handler
- More assertion
Update javadoc for CleartextHttp2ServerUpgradeHandler
Rename ClearTextHttp2ServerUpgradeHandler to CleartextHttp2ServerUpgradeHandler
Removed redundant code about configure pipeline
nit: PriorKnowledgeHandler
Removed Mockito.spy, investigate conn state instead
Add Http2UpgradeEvent
Check null of the constructor arguments
Rename Http2UpgradeEvent to PriorKnowledgeUpgradeEvent
Update unit test
Motivation
Http2StreamChannel ignores options of its parent channel when being created. That leads to surprising results when, for example, unpooled allocator could be silently replaced with pooled allocator (default setting).
Modification
Copy parent channel's options over to the Http2StreamChannel.
Result
Channel options are now consistent between Http2StreamChannel and its parent channel. Newly added test passes on this branch and fails on master. Fixes#6551.
Motivation:
We used some deprecated Mockito methods.
Modifications:
- Replace deprecated method usage
- Some cleanup
Result:
No more usage of deprecated Mockito methods. Fixes [#6482].
Motivation:
DefaultHttp2ConnectionDecoder#onSettingsRead processes the settings, and then sends a SETTINGS ACK to the remote peer. Processing the settings may result in frames which violate the previous settings being send to the remote peer. The remote peer will not apply the new settings until it has received the SETTINGS ACK, and therefore we may violate the settings from the remote peer's perspective and the connection will be shutdown.
Modifications:
- We should send the SETTINGS ACK before we process the settings to ensure the peer receives the SETTINGS ACK before other frames which assume the settings have already been applied
Result:
Fixes https://github.com/netty/netty/issues/6520.
Motivation:
The internal.hpack classes are no longer exposed in our public APIs and can be made package private in the http2 package.
Modifications:
- Make the hpack classes package private in the http2 package
Result:
Less APIs exposed as public.
Motivation:
The timeouts used in the Http2ConnectionRoundtripTest seems to be too low when leak-detection is enabled so we sometimes get failed tests due timeout.
Modifications:
Increase timeouts.
Result:
Fixes [#6442].
Motivation:
DefaultHttp2FrameWriter contains important logic for how http2
frame encode into binary, currently, netty had no unit test
just for DefaultHttp2FrameWriter.
Modifications:
Write DefaultHttp2FrameWriterTest to test DefaultHttp2FrameWriter
Result:
The coverage of DefaultHttp2FrameWriter is increased, and the
class was more reliable
Motivation:
Some unit HTTP/2 unit tests use LocalChannel. LocalChannel's doClose method will ensure any pending items in the queue will be released, but it may execute a Runnable on the peer's EventLoop to ensure the peer's queue is also cleaned up. The HTTP/2 unit tests close the event loop groups with no wait time so that unit tests will execute quickly, but if the doClose Runnable is in the EventLoop's queue it will not run and thus the items in the queue will not be released.
Modifications:
- Ensure all HTTP/2 unit tests which use LocalChannel wait for both client and server channels to be closed before closing the EventLoop.
Result:
Related to https://github.com/netty/netty/issues/5850.
Motivation:
DefaultHttp2FrameReader contains the logic for how it parsed the network
traffic from http2 client,
it also validate the content is legal or not.
So keep high coverage rate on it will increase the stability of api.
Modifications:
Add unit test on DefaultHttp2FrameReader
Result:
Coverage rate increased
Motivation:
We used various mocking frameworks. We should only use one...
Modifications:
Make usage of mocking framework consistent by only using Mockito.
Result:
Less dependencies and more consistent mocking usage.
Motivation:
codec-http2 couples the dependency tree state with the remainder of the stream state (Http2Stream). This makes implementing constraints where stream state and dependency tree state diverge in the RFC challenging. For example the RFC recommends retaining dependency tree state after a stream transitions to closed [1]. Dependency tree state can be exchanged on streams in IDLE. In practice clients may use stream IDs for the purpose of establishing QoS classes and therefore retaining this dependency tree state can be important to client perceived performance. It is difficult to limit the total amount of state we retain when stream state and dependency tree state is combined.
Modifications:
- Remove dependency tree, priority, and weight related items from public facing Http2Connection and Http2Stream APIs. This information is optional to track and depends on the flow controller implementation.
- Move all dependency tree, priority, and weight related code from DefaultHttp2Connection to WeightedFairQueueByteDistributor. This is currently the only place which cares about priority. We can pull out the dependency tree related code in the future if it is generally useful to expose for other implementations.
- DefaultHttp2Connection should explicitly limit the number of reserved streams now that IDLE streams are no longer created.
Result:
More compliant with the HTTP/2 RFC.
Fixes https://github.com/netty/netty/issues/6206.
[1] https://tools.ietf.org/html/rfc7540#section-5.3.4
Motivation:
If an HTTP/2 client writes data before the connection preface the peer will shutdown the socket. Depending on what is in the pipeline (SslHandler) may require different evaluation criteria to infer when the codec-http2 has written the connection preface on behalf of the client. This can lead to unnecessarily complexity and error prone/racy application code.
Modifications:
- Introduce a user event that is fired up the pipeline when codec-http2 writes the connection preface
Result:
Reliable mechanism for applications to use to know when connection preface has been written (related to https://github.com/netty/netty/issues/6272).
Motivation:
When An HTTP server is listening in plaintext mode, it doesn't have
a chance to negotiate "h2" in the tls handshake. HTTP 1 clients
that are not expecting an HTTP2 server will accidentally a request
that isn't an upgrade, which the HTTP/2 decoder will not
understand. The decoder treats the bytes as hex and adds them to
the error message.
These error messages are hard to understand by humans, and result
in extra, manual work to decode.
Modification:
If the first bytes of the request are not the preface, the decoder
will now see if they are an HTTP/1 request first. If so, the error
message will include the method and path of the original request in
the error message.
In case the path is long, the decoder will check up to the first
1024 bytes to see if it matches. This could be a DoS vector if
tons of bad requests or other garbage come in. A future optimization
would be to treat the first few bytes as an AsciiString and not do
any Charset decoding. ByteBuf.toCharSequence alludes to such an
optimization.
The code has been left simple for the time being.
Result:
Faster identification of errant HTTP requests.
Motivation:
2fd42cfc6b fixed a bug related to encoding headers but it also introduced a throws statement onto the Http2FrameWriter methods which write headers. This throws statement makes the API more verbose and is not necessary because we can communicate the failure in the ChannelFuture that is returned by these methods.
Modifications:
- Remove throws from all Http2FrameWriter methods.
Result:
Http2FrameWriter APIs do not propagate checked exceptions.
Motivation:
At rfc7540 5.5, it said that it's not permitted to send extension
frame in the middle of header block and need be treated as
protocol error
Modifications:
When received a extension frame, in netty it's called unknown frame,
will verify that is there an headersContinuation exists
Result:
When received a extension frame in the middle of header block,
will throw connection error and closed the connection
This prevents us from having the first request, that hasn't ack'ed the setting causing a GOAWAY when we'd would
be under the maxHeaderListSizeGoAway that would have been set after the settings ack.
Motivation:
If the HPACK Decoder detects that SETTINGS_MAX_HEADER_LIST_SIZE has been violated it aborts immediately and sends a RST_STREAM frame for what ever stream caused the issue. Because HPACK is stateful this means that the HPACK state may become out of sync between peers, and the issue won't be detected until the next headers frame. We should make a best effort to keep processing to keep the HPACK state in sync with our peer, or completely close the connection.
If the HPACK Encoder is configured to verify SETTINGS_MAX_HEADER_LIST_SIZE it checks the limit and encodes at the same time. This may result in modifying the HPACK local state but not sending the headers to the peer if SETTINGS_MAX_HEADER_LIST_SIZE is violated. This will also lead to an inconsistency in HPACK state that will be flagged at some later time.
Modifications:
- HPACK Decoder now has 2 levels of limits related to SETTINGS_MAX_HEADER_LIST_SIZE. The first will attempt to keep processing data and send a RST_STREAM after all data is processed. The second will send a GO_AWAY and close the entire connection.
- When the HPACK Encoder enforces SETTINGS_MAX_HEADER_LIST_SIZE it should not modify the HPACK state until the size has been checked.
- https://tools.ietf.org/html/rfc7540#section-6.5.2 states that the initial value of SETTINGS_MAX_HEADER_LIST_SIZE is "unlimited". We currently use 8k as a limit. We should honor the specifications default value so we don't unintentionally close a connection before the remote peer is aware of the local settings.
- Remove unnecessary object allocation in DefaultHttp2HeadersDecoder and DefaultHttp2HeadersEncoder.
Result:
Fixes https://github.com/netty/netty/issues/6209.
Motivation:
- Decoder#decodeULE128 has a bounds bug and cannot decode Integer.MAX_VALUE
- Decoder#decodeULE128 doesn't support values greater than can be represented with Java's int data type. This is a problem because there are cases that require at least unsigned 32 bits (max header table size).
- Decoder#decodeULE128 treats overflowing the data type and invalid input the same. This can be misleading when inspecting the error that is thrown.
- Encoder#encodeInteger doesn't support values greater than can be represented with Java's int data type. This is a problem because there are cases that require at least unsigned 32 bits (max header table size).
Modifications:
- Correct the above issues and add unit tests.
Result:
Fixes https://github.com/netty/netty/issues/6210.
Motivation:
Build failures have been observed with 2 second timeouts on the CI servers. We should make the timeouts longer to reduce false positive test failures due to tests timing out prematurely.
Modifications:
- Increase timeouts from 2 and 3 seconds to 5 seconds.
Result:
Less false positive test failures.
Motivation:
When DefaultHttp2Connection removes a stream it iterates over all children and adds them as children to the parent of the stream being removed. This process may remove elements from the child map while iterating without using the iterator's remove() method. This is generally unsafe and may result in an undefined iteration.
Modifications:
- We should use the Iterator's remove() method while iterating over the child map
Result:
Fixes https://github.com/netty/netty/issues/6163
Motivation:
Currently clients attempting to send headers that are too large recieve
a RST frame. This makes it harder than needed for implementations on top
of netty to handle this in a graceful way.
Modifications:
When the Decoder throws a StreamError of type FRAME_SIZE_ERROR, the
Http2ConnectionHandler will now attempt to send an Http2Header with
status 431 and endOfStream=true
Result:
Implementations now do not have to subclass parts of netty to handle
431s
Motivation:
The 2 second timeout to bootstrap the test can timeout on the build servers. We should increase the timeout so it is less likely under powered or over worked machines are less likely to generate false failures.
Modifications:
- HttpToHttp2ConnectionHandlerTest setup timeout changed from 2 to 5 seconds
Result:
Less false build failures.
Motivation:
LastInboundHandler maintains 2 queues which may contain the same data and tries to match these up when you read elements out of it. Because of this it can happen that you remove an element only out of one queue and so double free stuff later.
Modifications:
Just use one "queue" to store things.
Result:
Not possible to only remove things from one queue and so get into trouble later when release everything that sits in the handler.
Motivation:
When a LocalChannel is closed it is responsible to ensure all queued objects are released. When a LocalChannel is closed it will also close its peer channel. However in HTTP/2 unit tests we may not wait until all channels have completed the shutdown process before destroying the threads and exiting the test. This may mean buffers are GCed before they are released and be reported as a leak.
Modifications:
- In HTTP/2 tests when we use LocalChannel we should wait for all channels to close before exiting the test and cleaning up the associated EventLoopGroups.
Result:
More correct usage of LocalChannel in HTTP/2 tests.
Motiviation:
We used ReferenceCountUtil.releaseLater(...) in our tests which simplifies a bit the releasing of ReferenceCounted objects. The problem with this is that while it simplifies stuff it increase memory usage a lot as memory may not be freed up in a timely manner.
Modifications:
- Deprecate releaseLater(...)
- Remove usage of releaseLater(...) in tests.
Result:
Less memory needed to build netty while running the tests.
Motivation:
If a stream is not able to send any data (flow control window for the stream is exhausted) but has descendants who can send data then WeightedFairQueueByteDistributor may incorrectly modify the pseudo time and also double add the associated state to the parent's priority queue. The pseudo time should only be modified if a node is moved in the priority tree, and not if there happens to be no active streams in its descendent tree and a descendent is moved (e.g. removed from the tree because it wrote all data and the last data frame was EOS). Also the state objects for WeightedFairQueueByteDistributor should only appear once in any queue. If this condition is violated the pseudo time accounting would be biased at and assumptions in WeightedFairQueueByteDistributor would be invalidated.
Modifications:
- WeightedFairQueueByteDistributor#isActiveCountChangeForTree should not allow re-adding to the priority queue if we are currently processing a node in the distribution algorithm. The distribution algorithm will re-evaluate if the node should be re-added on the tail end of the recursion.
Result:
Fixes https://github.com/netty/netty/issues/5980
Motivation:
Currently it is not possible to have an Http/2 server send non default
initial settings to clients when doing the initial connection handshake
Modifications:
Add additional constructors to Http2Codec allowing users to specify the
initial settings to send to the client and apply locally
Result:
You can now specify non default initial settings
Motivation:
The SETTINGS_MAX_HEADER_LIST_SIZE limit, as enforced by the HPACK Encoder, should be a stream error and not apply to the whole connection.
Modifications:
Made the necessary changes for the exception to be of type StreamException.
Result:
A HEADERS frame exceeding the limit, only affects a specific stream.
Motivation:
The responsibility for retaining the settings values and enforcing the settings constraints is spread out in different areas of the code and may be initialized with different values than the default specified in the RFC. This should not be allowed by default and interfaces which are responsible for maintaining/enforcing settings state should clearly indicate the restrictions that they should only be set by the codec upon receipt of a SETTINGS ACK frame.
Modifications:
- Encoder, Decoder, and the Headers Encoder/Decoder no longer expose public constructors that allow the default settings to be changed.
- Http2HeadersDecoder#maxHeaderSize() exists to provide some bound when headers/continuation frames are being aggregated. However this is roughly the same as SETTINGS_MAX_HEADER_LIST_SIZE (besides the 32 byte octet for each header field) and can be used instead of attempting to keep the two independent values in sync.
- Encoding headers now enforces SETTINGS_MAX_HEADER_LIST_SIZE at the octect level. Previously the header encoder compared the number of header key/value pairs against SETTINGS_MAX_HEADER_LIST_SIZE instead of the number of octets (plus 32 bytes overhead).
- DefaultHttp2ConnectionDecoder#onData calls shouldIgnoreHeadersOrDataFrame but may swallow exceptions from this method. This means a STREAM_RST frame may not be sent when it should for an unknown stream and thus violate the RFC. The exception is no longer swallowed.
Result:
Default settings state is enforced and interfaces related to settings state are clarified.
