[Net] Fix IP address resolution incorrectly locking the main thread. #51199
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Outdated suggestions cannot be applied.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.
This seems to be a pretty old bug, older then originally reported (at least under certain circumstances).
The IP singleton uses a resolve queue so developers can queue hostnames for resolution in a separate while keeping the main thread unlocked (address-resolution OS functions are blocking, and could block for a long time in case of network disruption).
In most places though, the address resolution function was called with the mutex locked, causing other functions (querying status, queueing another hostname, ecc) to block until that resolution ended.
This commit ensures that all calls to OS address resolution are done with the mutex unlocked.
Fixes #51181 . Note: wile in the issue this is described as a regression, this is a very old issue (according to my tests), according to my tests on Linux, and can be found even to
2.1
. I suspect testing it is just tricky due to OS/router DNS caches.3.x
version here