This repository has been archived by the owner on Jan 23, 2023. It is now read-only.
Fix ClientWebSocket close handshake when using proxy #36928
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
When the client websocket was going through a proxy, it would sometimes transition to the 'Aborted' state and not the 'Closed' state after a successful closing handshake. I first opened this bug a year ago but finally was able to get back to it and root cause the problem. The problem only occured with the managed websocket. The .NET Framework did not have this problem. And some proxies didn't cause a problem with managed websocket (such as Fiddler). The root cause is a misinterpretation of RFC 6455, Section 7.1.1. This describes the behavior of how the websocket's TCP connection should be closed. The most graceful way is to wait for the server to initiate the close of the socket. In cases where it is taking too long to wait for the server to start the TCP close, then the client can start the TCP close of the socket. But no matter how the socket is finally closed, the websocket state should transition to 'Closed' as soon as a valid closing handshake was performed (i.e. close frames both sent and received). And this occurs before any logic for closing the TCP connection. The code in managed websocket has a timer to wait 1 second for the server to start a close. If the timer finishes, then the managed websocket calls an Abort() method to close the socket. This ends up transitioning the websocket to an 'Aborted' state which is incorrect. The state should only be moved to the 'Aborted' state if a closing handshake was not completed. I added a new test to support this change and moved the LoopbackProxyServer code to Common. Fixes #28027
|
/azp run |
|
Azure Pipelines successfully started running 4 pipeline(s). |
|
Outerloop failures not related to PR:
|
|
Innerloop failures not related to PR:
|
@davidsh, they are related. Those HttpListener failures are asserts due to this change. All PRs are failing as a result. |
Sigh. Ok. I'll get that fix/disabled asap. |
davidsh
added a commit
to davidsh/corefx
that referenced
this pull request
Apr 17, 2019
An HttpListener websocket test was failing after the change from PR dotnet#36928. That PR made changes to tighten up the transition to the Closed state after the closing handshake had completed. That PR missed some additional changes needed especially in cases where a server (such as loopback) sends the close frame almost concurrently with the client sending the close frame. This PR fixes the close frame handshake logic and also makes some optimizations in cases where the websocket has already transitioned to the Aborted state during the closing handshake. In that case, the websocket should not wait for a close frame to be received but should simply proceed to closing the connection. Fixes #36963
davidsh
added a commit
that referenced
this pull request
Apr 18, 2019
An HttpListener websocket test was failing after the change from PR #36928. That PR made changes to tighten up the transition to the Closed state after the closing handshake had completed. That PR missed some additional changes needed especially in cases where a server (such as loopback) sends the close frame almost concurrently with the client sending the close frame. This PR fixes the close frame handshake logic and also makes some optimizations in cases where the websocket has already transitioned to the Aborted state during the closing handshake. In that case, the websocket should not wait for a close frame to be received but should simply proceed to closing the connection. Fixes #36963
picenka21
pushed a commit
to picenka21/runtime
that referenced
this pull request
Feb 18, 2022
…6928) When the client websocket was going through a proxy, it would sometimes transition to the 'Aborted' state and not the 'Closed' state after a successful closing handshake. I first opened this bug a year ago but finally was able to get back to it and root cause the problem. The problem only occured with the managed websocket. The .NET Framework did not have this problem. And some proxies didn't cause a problem with managed websocket (such as Fiddler). The root cause is a misinterpretation of RFC 6455, Section 7.1.1. This describes the behavior of how the websocket's TCP connection should be closed. The most graceful way is to wait for the server to initiate the close of the socket. In cases where it is taking too long to wait for the server to start the TCP close, then the client can start the TCP close of the socket. But no matter how the socket is finally closed, the websocket state should transition to 'Closed' as soon as a valid closing handshake was performed (i.e. close frames both sent and received). And this occurs before any logic for closing the TCP connection. The code in managed websocket has a timer to wait 1 second for the server to start a close. If the timer finishes, then the managed websocket calls an Abort() method to close the socket. This ends up transitioning the websocket to an 'Aborted' state which is incorrect. The state should only be moved to the 'Aborted' state if a closing handshake was not completed. I added a new test to support this change and moved the LoopbackProxyServer code to Common. Fixes dotnet/corefx#28027 Commit migrated from dotnet/corefx@0c9684c
picenka21
pushed a commit
to picenka21/runtime
that referenced
this pull request
Feb 18, 2022
An HttpListener websocket test was failing after the change from PR dotnet/corefx#36928. That PR made changes to tighten up the transition to the Closed state after the closing handshake had completed. That PR missed some additional changes needed especially in cases where a server (such as loopback) sends the close frame almost concurrently with the client sending the close frame. This PR fixes the close frame handshake logic and also makes some optimizations in cases where the websocket has already transitioned to the Aborted state during the closing handshake. In that case, the websocket should not wait for a close frame to be received but should simply proceed to closing the connection. Fixes dotnet/corefx#36963 Commit migrated from dotnet/corefx@1824535
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When the client websocket was going through a proxy, it would sometimes transition to the
'Aborted' state and not the 'Closed' state after a successful closing handshake.
I first opened this bug a year ago but finally was able to get back to it and root cause the
problem. The problem only occured with the managed websocket. The .NET Framework did not have
this problem. And some proxies didn't cause a problem with managed websocket (such as Fiddler).
The root cause is a misinterpretation of RFC 6455, Section 7.1.1. This describes the behavior
of how the websocket's TCP connection should be closed. The most graceful way is to wait for
the server to initiate the close of the socket. In cases where it is taking too long to wait
for the server to start the TCP close, then the client can start the TCP close of the socket.
But no matter how the socket is finally closed, the websocket state should transition to 'Closed'
as soon as a valid closing handshake was performed (i.e. close frames both sent and received).
And this occurs before any logic for closing the TCP connection.
The code in managed websocket has a timer to wait 1 second for the server to start a close. If
the timer finishes, then the managed websocket calls an Abort() method to close the socket. This
ends up transitioning the websocket to an 'Aborted' state which is incorrect. The state should
only be moved to the 'Aborted' state if a closing handshake was not completed.
I added a new test to support this change and moved the LoopbackProxyServer code to Common.
Fixes #28027