Motivation:
This is small fixes for #10453 PR according @njhill and @normanmaurer conversation.
Modification:
Simple refactor and takes into account remainder when calculate size.
Result:
Behavior is correct
Motivation:
AbstractDiskHttpData#getChunk opens and closes fileChannel every time when it is invoked,
as a result the uploaded file is corrupted. This is a regression caused by #10270.
Modifications:
- Close the fileChannel only if everything was read or an exception is thrown
- Add unit test
Result:
AbstractDiskHttpData#getChunk closes fileChannel only if everything was read or an exception is thrown
Motivation:
Regression appeared after making changes in fix#10360 .
The main problem here that `buffer.getBytes(buffer.readerIndex(), fileChannel, fileChannel.position(), localsize)`
doesn't change channel position after writes.
Modification:
Manually set position according to the written bytes.
Result:
Fixes#10449 .
Motivation:
Since https://github.com/netty/netty/pull/9865 (Netty 4.1.44) the
default behavior of the HttpObjectDecoder has been to reject any HTTP
message that is found to have multiple Content-Length headers when
decoding. This behavior is well-justified as per the risks outlined in
https://github.com/netty/netty/issues/9861, however, we can see from the
cited RFC section that there are multiple possible options offered for
responding to this scenario:
> If a message is received that has multiple Content-Length header
> fields with field-values consisting of the same decimal value, or a
> single Content-Length header field with a field value containing a
> list of identical decimal values (e.g., "Content-Length: 42, 42"),
> indicating that duplicate Content-Length header fields have been
> generated or combined by an upstream message processor, then the
> recipient MUST either reject the message as invalid or replace the
> duplicated field-values with a single valid Content-Length field
> containing that decimal value prior to determining the message body
> length or forwarding the message.
https://tools.ietf.org/html/rfc7230#section-3.3.2
Netty opted for the first option (rejecting as invalid), which seems
like the safest, but the second option (replacing duplicate values with
a single value) is also valid behavior.
Modifications:
* Introduce "allowDuplicateContentLengths" parameter to
HttpObjectDecoder (defaulting to false).
* When set to true, will allow multiple Content-Length headers only if
they are all the same value. The duplicated field-values will be
replaced with a single valid Content-Length field.
* Add new parameterized test class for testing different variations of
multiple Content-Length headers.
Result:
This is a backwards-compatible change with no functional change to the
existing behavior.
Note that the existing logic would result in NumberFormatExceptions
for header values like "Content-Length: 42, 42". The new logic correctly
reports these as IllegalArgumentException with the proper error message.
Additionally note that this behavior is only applied to HTTP/1.1, but I
suspect that we may want to expand that to include HTTP/1.0 as well...
That behavior is not modified here to minimize the scope of this change.
Motivation:
AbstractDiskHttpData may cause a memory leak when a CompositeByteBuf is used. This happened because we may call copy() but actually never release the newly created ByteBuf.
Modifications:
- Remove copy() call and just use ByteBuf.getBytes(...) which will internally handle the writing to the FileChannel without any extra copies that need to be released later on.
- Add unit test
Result:
Fixes https://github.com/netty/netty/issues/10354
Motivation
HttpObjectDecoder and its associated classes make frequent use of
default values for maxInitialLineLength, maxHeaderSize, maxChunkSize,
etc. Today, these defaults are defined in-line in constructors and
duplicated across many classes. This repetition is more prone to error
and inconsistencies.
Furthermore, due to the current lack of builder support, if a user wants
to change just one of these values (e.g., maxHeaderSize), they are also
required to know and repeat the other default values (e.g.,
maxInitialLineLength and maxChunkSize).
The primary motivation for this change is as we are considering adding
another constructor parameter (for multiple content length behavior),
appending this parameter may require some users to have prior knowledge
of the default initialBufferSize, and it would be cleaner to allow them
to reference the default constant.
Modifications
* Consolidate the HttpObjectDecoder default values into public constants
* Reference these constants where possible
Result
No functional change. Additional telescoping constructors will be easier
and safer to write. Users may have an easier experience changing single
parameters.
Motivation:
`containsValue()` will check if there are multiple values defined in the specific header name, we need to use this method instead of `contains()` for the `Transfer-Encoding` header to cover the case that multiple values defined, like: `Transfer-Encoding: gzip, chunked`
Modification:
Change from `contains()` to `containsValue()` in `HttpUtil.isTransferEncodingChunked()` method.
Result:
Fixes#10320
Motivation:
Currently we passing custom websocket handshaker response headers to a `WebSocketServerHandshaker` but they can contain a reserved headers (e.g. Connection, Upgrade, Sec-Websocket-Accept) what lead to duplication because we use response.headers().add(..) instead of response.headers().set(..).
Modification:
In each `WebSocketServerHandshaker00`, ... `WebSocketServerHandshaker13` implementation replace the method add(..) to set(..) for reserved response headers.
Result:
Less error-prone
Motivation:
`FileChannel.read()` may throw an IOException. We must deal with this in case of the occurrence of `I/O` error.
Modification:
Place the `FileChannel.read()` method call in the `try-finally` block.
Result:
Advoid fd leak.
Co-authored-by: Norman Maurer <norman_maurer@apple.com>
Motivations
-----------
HttpPostStandardRequestDecoder was changed in 4.1.50 to provide its own
ByteBuf UrlDecoder. Prior to this change, it was using the decodeComponent
method from QueryStringDecoder which decoded + characters to
whitespaces. This behavior needs to be preserved to maintain backward
compatibility.
Modifications
-------------
Changed HttpPostStandardRequestDecoder to detect + bytes and decode them
toe whitespaces. Added a test.
Results
-------
Addresses issues#10284
Motivation:
We cant reuse the ChannelPromise as it will cause an error when trying to ful-fill it multiple times.
Modifications:
- Use a new promise and chain it with the old one
- Add unit test
Result:
Fixes https://github.com/netty/netty/issues/10240
Motivation:
We should not use Unpooled to allocate buffers if possible to ensure we can make use of pooling etc.
Modifications:
- Only allocate a buffer if really needed
- Use the ByteBufAllocator of the offered ByteBuf
- Ensure we not use buffer.copy() but explicitly allocate a buffer and then copy into it to not hit the limit of maxCapacity()
Result:
Improve allocations
Motivation:
We need to release all ByteBufs that we allocate to prevent leaks. We missed to release the ByteBufs that are used to aggregate in two cases
Modifications:
Add release() calls
Result:
No more memory leak in HttpPostMultipartRequestDecoder
Motivations
-----------
There is no need to copy the "offered" ByteBuf in HttpPostRequestDecoder
when the first HttpContent ByteBuf is also the last (LastHttpContent) as
the full content can immediately be decoded. No extra bookeeping needed.
Modifications
-------------
HttpPostMultipartRequestDecoder
- Retain the first ByteBuf when it is both the first HttpContent offered
to the decoder and is also LastHttpContent.
- Retain slices of the final buffers values
Results
-------
ByteBufs of FullHttpMessage decoded by HttpPostRequestDecoder are no longer
unnecessarily copied. Attributes are extracted as retained slices when
the content is multi-part. Non-multi-part content continues to return
Unpooled buffers.
Partially addresses issue #10200
Motivation:
`AbstractHttpData.checkSize` may throw an IOException if we set the max size limit via `AbstractHttpData.setMaxSize`. However, if this exception happens, the `AbstractDiskHttpData.file` and the `AbstractHttpData.size` are still be modified. In other words, it may break the failure atomicity here.
Modification:
Just move up the size check.
Result:
Keep the failure atomicity even if `AbstractHttpData.checkSize` fails.
Motivation:
An unexpected IOException may be thrown from `FileChannel.force`. If it happens, the `FileChannel.close` may not be invoked.
Modification:
Place the `FileChannel.close` in a finally block.
Result:
Avoid fd leak.
Motivation:
`RandomAccessFile.setLength` may throw an IOException. We must deal with this in case of the occurrence of `I/O` error.
Modification:
Place the `RandomAccessFile.setLength` method call in the `try-finally` block.
Result:
Avoid fd leak.
Motivation:
`FileChannel.force` may throw an IOException. A fd leak may happen here.
Modification:
Close the fileChannel in a finally block.
Result:
Avoid fd leak.
Motivation:
An `IOException` may be thrown from `FileChannel.write` or `FileChannel.force`, and cause the fd leak.
Modification:
Close the file in a finally block.
Result:
Avoid fd leak.
Motivation:
An exception may occur between ByteBuf's allocation and release. So I think it's a good idea to place the release operation in a finally block.
Modification:
Release the temporary ByteBuf in finally blocks.
Result:
Avoid ByteBuf leak.
Motivation:
An IOException may be thrown from FileChannel.read, and cause the fd leak.
Modification:
Close the file when IOException occurs.
Result:
Avoid fd leak.
Motivation:
currently (http) disk based attributes or uploads are globally configured in a single directory and can also only globally be deleted on exit or not. it does not fit well multiple cases, in particular the case you have multiple servers in the same JVM.
Modification:
make it configurable per attribute/fileupload.
Result:
This PR duplicates Disk* constructor to add basedir and deleteonexit parameters and wires it in default http daa factory.
Motivation:
An IOException may be thrown from InputStream.read or checkSize method, and cause the ByteBuf leak.
Modification:
Release the ByteBuf when IOException occurs.
Result:
Avoid ByteBuf leak.
Motivation:
At the moment HttpObjectDecoder does not limit the number of controls chars. These should be counted with the initial line and so a limit should be exposed
Modifications:
- Change LineParser to also be used when skipping control chars and so enforce a limit
- Add various tests to ensure that limit is enforced
Result:
Fixes https://github.com/netty/netty/issues/10111
Motivation:
In our WebSocketClientHandshaker* implementations we "hardcode" the version number to use. This is error-prone, we should better use the WebSocketVersion so we dont need to maintain the value multiple times. Beside this we can also use an AsciiString to improve performance
Modifications:
- Use WebSocketVersion.toAsciiString
Result:
Less stuff to maintain and small performance win
Motivation:
The user may need to provide a specific HOST header. We should not override it when specified during handshake.
Modifications:
Check if a custom HOST header is already provided by the user and if so dont override it
Result:
Fixes https://github.com/netty/netty/issues/10101
Motivation:
Netty currently does not support the SameSite attribute for response cookies (see issue #8161 for discussion).
Modifications:
The attribute has been added to the DefaultCookie class as a quick fix since adding new methods to the Cookie interface would be backwards-incompatible.
ServerCookieEncoder and ClientCookieDecoder have been updated accordingly to process this value. No validation for allowed values (Lax, None, Strict) has been implemented.
Result:
Response cookies with the SameSite attribute set can be read or written by Netty.
Co-authored-by: David Latorre <a-dlatorre@hotels.com>
Motivation:
WebSocketClientHandshaker#upgradeUrl doesn't comperly compute relative url when path is empty and produces url such as `?access_token=foo` instead of `/?access_token=foo`.
Modifications:
* fix WebSocketClientHandshaker#upgradeUrl
* add tests for urls without path, with and without query
Result:
WebSocketClientHandshaker properly connects to url without path.
Motivation:
I am receiving a mutlipart/form_data upload from postman. The filename contains Chinese, and so some invalid chars. We should ensure all of these are removed before trying to decode.
Modification:
Ensure all invalid characters are removed
Result:
Fixes#10087
Co-authored-by: liming.zhang <liming.zhang@luckincoffee.com>
Motivation:
When the HttpContentCompressor is put on an alternate EventExecutor, the order of events should be
preserved so that the compressed content is correctly created.
Modifications:
- checking that the executor in the ChannelHandlerContext is the same executor as the current executor when evaluating if the handler should be skipped.
- Add unit test
Result:
Fixes https://github.com/netty/netty/issues/10067
Co-authored-by: Norman Maurer <norman_maurer@apple.com>
Motivation:
Our parsing of the initial line / http headers did treat some characters as separators which should better trigger an exception during parsing.
Modifications:
- Tighten up parsing of the inital line by follow recommentation of RFC7230
- Restrict separators to OWS for http headers
- Add unit test
Result:
Stricter parsing of HTTP1
Motivation:
We did had some System.out.println(...) call in a test which seems to be some left-over from debugging.
Modifications:
Remove System.out.println(...)
Result:
Code cleanup
Motivation
As part of a recent commit for issue
https://github.com/netty/netty/issues/9861 the HttpObjectDecoder was
changed to throw an IllegalArgumentException (and produce a failed
decoder result) when decoding a message with both "Transfer-Encoding:
chunked" and "Content-Length".
While it seems correct for Netty to try to sanitize these types of
messages, the spec explicitly mentions that the Content-Length header
should be *removed* in this scenario.
Both Nginx 1.15.9 and Tomcat 9.0.31 also opt to remove the header:
b693d7c198/java/org/apache/coyote/http11/Http11Processor.java (L747-L755)0ad4393e30/src/http/ngx_http_request.c (L1946-L1953)
Modifications
* Change the default behavior from throwing an IllegalArgumentException
to removing the "Content-Length" header
* Extract the behavior to a new protected method,
handleChunkedEncodingWithContentLength(), that can be overridden to
change this behavior (or capture metrics)
Result
Messages of this nature will now be successfully decoded and have their
"Content-Length" header removed, rather than creating invalid messages
(decoder result failures). Users will be allowed to override and
configure this behavior.
Motivation:
Need tests to ensure that CVE-2020-7238 is fixed.
Modifications:
Added two test cases into HttpRequestDecoderTest which check that
no whitespace is allowed before the Transfer-Encoding header.
Result:
Improved test coverage for #9861
Motivation:
Avoid allocation of default static `WebSocketServerProtocolConfig` and `WebSocketClientProtocolConfig` configs. Prefer compile time constants instead.
Modification:
Static field with config object replaced with constructor with default fields.
Result:
No more default config allocation and static field for it. Compile time variables used instead.