api: replace use of DevCreateObjectQuery with CM_Register_Notifcation… #7
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
… and ensure that the device interface list is explicitly queried immediately after notification registration Tailscale has several open issues concerning WinTun adapter installations that hang indefinitely, stuck waiting on an event that never triggers. Based on analysis of Tailscale logs and examination of WinTun's adapter installation code, I believe we are seeing a race condition that only manifests under specific timing scenarios that are relatively infrequent for most users, but are common for some. It is hard to reason about DevCreateObjectQuery because it is an undocumented API. Based upon my reading of the existing code and examination of existing documentation, it looks like CM_Register_Notification is the documented counterpart. Its documentation makes it pretty clear[1] that these notifications are essentially edge-triggered; we need to call CM_Get_Device_Interface_List immediately after registering the notification to ensure that the arrival event has not already happened. This patch replaces DevCreateObjectQuery with CM_Register_Notification because using a documented API gives us a better understanding of the semantics involved and makes this code easier to reason about. Then we add a call to AdapterGetDeviceObjectFileName to handle the aforementioned explicit check, thus ensuring that we are not waiting on an event that might have occurred prior to registering for event notifications. [1] https://learn.microsoft.com/en-us/windows/win32/api/cfgmgr32/nf-cfgmgr32-cm_register_notification#remarks Signed-off-by: Aaron Klotz <[email protected]>
|
Hey @dblohm7 👋 . Were you ever able to somehow reproduce this scenario? |
|
I started to use WireGuard thoes months, and calling WireGuardCreateAdapter might fail by accident. The error message is: "Timed out waiting for device query". It seems like the device create event is missing and result in WaitForSingleObject interface timeout. Most times recall WireGuardCreateAdapter will success, but rarely recall didn't work and I should restart the OS. It's hard to reproduce this problem Cause I don't know how it happened..... @dblohm7 Am I caught in the same situation which you're trying to fix? If so I'd like to pull this commit and see whether it works. |
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.
… and ensure that the device interface list is explicitly queried immediately after notification registration
Tailscale has several open issues concerning WinTun adapter installations that hang indefinitely, stuck waiting on an event that never triggers. Based on analysis of Tailscale logs and examination of WinTun's adapter installation code, I believe we are seeing a race condition that only manifests under specific timing scenarios that are relatively infrequent for most users, but are common for some.
It is hard to reason about DevCreateObjectQuery because it is an undocumented API. Based upon my reading of the existing code and examination of existing documentation, it looks like CM_Register_Notification is the documented counterpart. Its documentation makes it pretty clear[1] that these notifications are essentially edge-triggered; we need to call CM_Get_Device_Interface_List immediately after registering the notification to ensure that the arrival event has not already happened.
This patch replaces DevCreateObjectQuery with CM_Register_Notification because using a documented API gives us a better understanding of the semantics involved and makes this code easier to reason about. Then we add a call to AdapterGetDeviceObjectFileName to handle the aforementioned explicit check, thus ensuring that we are not waiting on an event that might have occurred prior to registering for event notifications.
[1] https://learn.microsoft.com/en-us/windows/win32/api/cfgmgr32/nf-cfgmgr32-cm_register_notification#remarks