Fix buffer leak in HTTP and FCGI when the server is being shut down while there are in-flight requests
Signed-off-by: Ludovic Orban <lorban@bitronix.be>
#10226
- fix HttpClientIdleTimeoutTest to wait for server's idle timeout before checking for leaks
- improve HttpClientIdleTimeoutTest by making it upload some content
- fix FCGI server leak caused by idle timeout
- fix H3 server leak caused by idle timeout
Signed-off-by: Ludovic Orban <lorban@bitronix.be>
Fixes#10234
* Introduced transient failures in reads where a failure chunk has last=false.
* Transient failure now do not fail the handler callback.
* Improve eeN ContentProducer to more carefully assert transient and terminal errors + enable HttpInputIntegrationTest
* Do not add connection: close to the response when the error is transient
* Rework ChunksContentSource to support null chunks
* Added tests to verify the new transient failure cases
* Review all code that handles failure, and handling correctly transient failure, either by making them fatal, and/or by failing Content.Source.
Signed-off-by: Ludovic Orban <lorban@bitronix.be>
Signed-off-by: Olivier Lamy <olamy@apache.org>
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
Co-authored-by: Ludovic Orban <lorban@bitronix.be>
Co-authored-by: Olivier Lamy <olamy@apache.org>
Co-authored-by: Joakim Erdfelt <joakim.erdfelt@gmail.com>
Co-authored-by: Chad Wilson <chadw@thoughtworks.com>
Co-authored-by: Simone Bordet <simone.bordet@gmail.com>
The local/remote SocketAddress is cached within the ConnectionMetaData or Connection instance, so that any changes are not visible during the request lifetime.
Ensure that all server Connection types respect HttpConfiguration#getLocalAddress and that it is not implemented only in servlet layer
Avoid DNS resolution.
---------
Co-authored-by: Joakim Erdfelt <joakim.erdfelt@gmail.com>
Introduced CompletableResponseListener, a replacement for FutureResponseListener that uses better APIs based on CompletableFuture.
Deprecated FutureResponseListener and replaced its usages.
Updated documentation.
Added tests for zipped request content.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
Simplified the code, removing all leftover cruft present since FCGI was multiplexed.
Now the FCGI implementation is very similar to HTTP1.
Made HttpConnectionOverFCGI.channel final so that it cannot NPE anymore.
The client now properly handling server-side connection closes.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
* Remove usage of instanceof Content.Chunk.Error
* Updated AsyncContent to accept a transient failures
* Updated AsyncContent to accept a transient failure with inputstream
* IdleTimeout review
+ pass TimeoutException through all APIs
+ HttpConnection now passes on TimeoutException to HttpChannel.onFailure
* More ServerTests for idletimeout
* Recreated a ServerTimeoutsTest for multiple transports
* more robust tests
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
* merged work from @sbordet and @gregw
* Various improvements to CyclicTimeouts.
* Improved reset of the earliest timeout before iteration.
* Removed check for getExpireNanoTime() == -1, since it's a valid value.
* When onExpired(Expirable) returns false, the Expirable should arrange to move its timeout in the future.
* fix keystore to please BoringSSL + use correct temp path
Signed-off-by: Ludovic Orban <lorban@bitronix.be>
* Fixed ErrorResponseAndCallback succeeded() and failed() to call super.failed() in all cases to complete the wrapped callback.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
* Revert "Fixed ErrorResponseAndCallback succeeded() and failed() to call super.failed() in all cases to complete the wrapped callback."
This reverts commit 5ac57c13e0.
* WIP idleTimeout
* WIP idleTimeout
* Added context wrapper for idle timeout listener
* updates from review
---------
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
Signed-off-by: Ludovic Orban <lorban@bitronix.be>
Co-authored-by: Simone Bordet <simone.bordet@gmail.com>
Co-authored-by: Ludovic Orban <lorban@bitronix.be>
* Fixed FCGI content parsing.
WordPress sends the initial content in the same frame as the response headers.
StreamContentParser was receiving the frame, delegating to HttpParser to parse the headers,
but the buffer was still containing some content, that was lost.
Now the content after the headers is correctly retained.
For the direct FastCGI APIs, now a read-only version of the buffer is passed to the listener,
and the original buffer is consumed as the caller expects.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
* Optimize Content-Length: 0 handling
Create and use a pre-encoded HttpFields.CONTENT_LENGTH_0 constant
Used the constant in more places
renamed the `putLongField` methods of HttpFields.Mutable to just `put`
Fixed wrong docs examples
* Removed unnecessary constructors from MetaData, MetaData.Request and MetaData.Response.
* Removed MetaData.Request.getURIString() (available as getHttpURI().toString()).
* Renamed MetaData.getFields() -> getHttpFields(), as they can be headers or trailers.
* Renamed MetaData.Request.getURI() -> getHttpURI().
* Normalized handling of contentLength, now always -1 (rather than Long.MIN_VALUE) if unknown.
* Permutated MetaData.Response constructor parameters to be consistent with MetaData.Request.
* MetaData.Request's method and httpURI must be non-null.
Signed-off-by: Simone Bordet <simone.bordet@gmail.com>
Convert class BadMessageException to a HttpMessage.RuntimeException to allows different types of HttpExceptions exceptions.
This follows the pattern of the QuietException interface.
Olivier Lamy
Joakim Erdfelt
Ludovic Orban
Greg Wilkins
Simone Bordet
Lachlan Roberts