Queue macOS input reports so that large responses aren't dropped #84
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
DennisDyallo
approved these changes
May 8, 2024
| if (_readHandle.IsAllocated) | ||
| { | ||
| _readHandle.Free(); | ||
| } | ||
|
|
||
| if (_pinnedReportsQueue.IsAllocated) | ||
| { | ||
| _pinnedReportsQueue.Free(); |
This addresses a bug where if a YubiKey command would result in multiple input reports being generated, we would get the callbacks but drop the results on the floor. This would generally result in the runloop timing out because we already had received the report via the callback but had no way of caching it.
GregDomzalski
force-pushed
the
bugfix/macos-hid-callbacks
branch
from
b8eefbc to
06edf90
Compare
Test Results: Windows 2 files 2 suites 2s ⏱️ Results for commit 5c10887. ♻️ This comment has been updated with latest results. |
Test Results: Ubuntu 2 files 2 suites 5s ⏱️ Results for commit 5c10887. ♻️ This comment has been updated with latest results. |
Test Results: MacOS 2 files 2 suites 1s ⏱️ Results for commit 5c10887. ♻️ This comment has been updated with latest results. |
Minimum allowed line rate is |
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Description
This has been a longstanding bug in the SDK that has been haunting me since the FIDO device code was originally written. Apple's documentation on how async input reports are intended to work are about as clear as mud. But with some extra (temporary) debug logging I was able to finally reproduce the issue with enough visibility into what was actually happening. The implementation used by our own
libfido2library now makes a lot more sense as well.Fundamentally, we were just using the
IOHIDDeviceRegisterInputReportCallbackand its corresponding callback incorrectly. The read buffer that we supply during the registration is just meant to be a place for macOS to cache the report data... I guess? The intended use still is not super clear to me.But what was clear from the logs is that the report callback was getting called faster than we were calling GetReport to read said reports. And since we previously did not queue up any of the reports until they were read, it resulted in reports either getting dropped, or arriving before we ran the IO runloop. That would usually result in the
CFRunLoopRunInModecall inside ofGetReportto timeout. Not because the YubiKey was taking a long time, but because the report already came and went and we missed it.So now, we pass a
ConcurrentQueueinto the native callback via aGCHandle. We queue the report that we see in the callback so that whenGetReportis eventually called, we will have the report. We first attempt to read straight from the queue, and if there's nothing there, we run the runloop which should then cause macOS to read from the YubiKey. This behavior now much more closely matcheslibfido2.Fixes (internal issue)
Type of change
Please delete options that are not relevant.
How has this been tested?
Ran some local FIDO2 tests as well as real-world use cases using our internally dependent software.
Test configuration:
Checklist:
dotnet formatto format my code