Fixes#4971 - Simplify Connection.upgradeFrom()/upgradeTo().
Now the upgrade-from connection produces a "floating" buffer
(not belonging to a pool), so that it can release the original buffer.
The upgrade-to connection is free to copy or store this "floating" buffer.
Strengthened ByteBufferPool behavior when releasing non-pooled
ByteBuffers: the buffer is now discarded.
Updated javadocs and all implementations.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
Fixes#4967 - Possible buffer corruption in HTTP/2 session failures
Partially reverted the changes introduced in #4855, because they
were working only when sends were synchronous.
Introduced ByteBufferPool.remove(ByteBuffer) to fix the issue.
Now when a concurrent failure happens while frames are being
generated or sent, the buffer is discarded instead of being
recycled, therefore resolving the buffer corruption.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
* Issue #4965 - WINDOW_UPDATE for locally failed stream should not close the HTTP/2 session.
Improved HTTP2Session.onWindowUpdate() code to correctly check whether
the stream is already closed, and if so, just drop the WINDOW_UPDATE.
Refactored onResetForUnknownStream() to base class.
Other small refactorings to improve logging.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
* Fixes#4855 - Occasional h2spec failures on CI
In case of bad usage of the HTTP/2 API, we don't want to close()
the stream but just fail the callback, because the stream
may be performing actions triggered by a legit API usage.
In case of a call to `AsyncListener.onError()`, applications may decide to call
AsyncContext.complete() and that would be a correct usage of the Servlet API.
This case was not well handled and was wrongly producing a WARN log with an
`IllegalStateException`.
Completely rewritten `HttpTransportOverHTTP2.TransportCallback`.
The rewrite handles correctly asynchronous failures that now are executed
sequentially (and not concurrently) with writes.
If a write is in progress, the failure will just change the state and at the
end of the write a check on the state will determine what actions to take.
A session failure is now handled in HTTP2Session by first failing all the
streams - which notifies the Stream.Listeners - and then failing the session
- which notifies the Session.Listener.
The stream failures are executed concurrently by dispatching each one to a
different thread; this means that the stream failure callbacks are executed
concurrently (likely sending RST_STREAM frames).
The session failure callback is completed only when all the stream failure
callbacks have completed, to ensure that a GOAWAY frame is processed after
all the RST_STREAM frames.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
- ArrayList contains() + add() is faster than HashSet add() for small collections
- A heap allocation of the iterator is required when iterating HashSet while iterating ArrayList can do with a stack allocation
Signed-off-by: Ludovic Orban <lorban@bitronix.be>
Introduced RateControl.Factory to create instances of RateControl
for each connection.
Modified relevant XML files and added distribution test for h2.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
Invalid HTTP/2 headers are now causing an error rather than being ignored.
HTTP2Flusher now catches HpackException.StreamException and generates a
RST_STREAM frame, rather than just closing the connection.
Modified HpackEncoder to throw HpackException in case of encoding failure.
Introduced HpackEncoder.validateEncoding (defaults true) so validation of
the headers can be disabled (useful for tests).
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
Introduced a Response.DemandedContentListener to explicitly separate
the will to request more content from the notification that the content
has been consumed.
Updated all transports to follow the new semantic: rather than waiting
for the callback to complete before delivering more content, now they
wait for the demand to be positive to deliver more content.
Since now the content may be unconsumed but there can be more demand,
all transport implementation had to be changed to use RetainableByteBuffer
to retain content buffers that were not consumed.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
Ludovic Orban
Joakim Erdfelt
Simone Bordet
Jesse Wilson
olivier lamy
Lachlan Roberts