Repository navigation
[release/9.0-staging] Link peer's X509 stack handle to parent SSL safe handle - #115380
jozkee merged 7 commits into
Conversation
Sorry, something went wrong.
…aphy.Native/Interop.Ssl.cs Co-authored-by: Kevin Jones <vcsjones@github.com>
|
@rzikm friendly reminder that code complete is on Monday May 12th (2:00 PM Pacific) for the June Release. If you'd like to get this change included in that release, please get a Tactics approval and merge this PR before the deadline. |
Sorry, something went wrong.
This reverts commit abf96c8.
|
Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones |
Sorry, something went wrong.
|
Reliability problem (crash) reported by 2 customers. Worth a fix. |
Sorry, something went wrong.
wfurt
left a comment
There was a problem hiding this comment.
LGTM. should we port the test as well for consistency?
Sorry, something went wrong.
bartonjs
left a comment
There was a problem hiding this comment.
I agree with Tomas, seems like the test should have been included.
Sorry, something went wrong.
* Defer RemoteCertificate assignment after X509 Chain build * Add comment
* Fix SslStreamDisposeTest.Dispose_ParallelWithHandshake_ThrowsODE test failures * Fix build
* [Test Failure] SslStreamDisposeTest.Dispose_ParallelWithHandshake_ThrowsODE on Unix Fixes #113833 * fixup! [Test Failure] SslStreamDisposeTest.Dispose_ParallelWithHandshake_ThrowsODE on Unix Fixes #113833 * fixup! fixup! [Test Failure] SslStreamDisposeTest.Dispose_ParallelWithHandshake_ThrowsODE on Unix Fixes #113833 * Update src/libraries/System.Net.Security/tests/FunctionalTests/SslStreamDisposeTest.cs Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Fix build --------- Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
|
The test in the original PR was flaky and uncovered other issues, so I had to include more PR (mostli test fixes). It should be good now. |
Sorry, something went wrong.
|
@rzikm reminder that servicing code complete for 9.0.7 is today @ 3PM PST. |
Sorry, something went wrong.
|
/ba-g pending leg hung with "Agents can't recover from restarts" but similar ones finished. |
Sorry, something went wrong.
|
Approved via email |
Sorry, something went wrong.
Backport of #113124 to release/9.0-staging
Fixes #109689
/cc @rzikm
Customer Impact
Reported by 2 customers. They experience process crashes when using specific pattern with short HTTP request timeout.
The crash happens when they dispose SslStream concurrently with an ongoing handshake operation on Linux. The concurrent dispose may occur when using HttpClient with very short timeout settings.
The fix is to properly use interior SafeHandle.
Regression
No -- the scenario is not too common and the timing in TLS 1.3 might be different from previous TLS versions, affecting chances of the problem happening.
Testing
Successfully reproduced on main with modified Runtime (artificial sleeps in code to increase likelihood of the race condition).
We lost customer repro environment after they updated OS and OpenSSL versions - we cannot validate on their side.
Risk
Low. The change is well contained and uses existing patterns.