Diff below changes when QH are added and removed to/from the async list.
It allows to cancel a transfer by simply unlinking its QH.
This should be enough to prevent the race in ehci_abort_xfer() resulting
in the infamous ehci_device_clear_toggle() panic.
I'd appreciate if people could test
unning some older OpenBSD version).
The panic is triggered when GoldenCheetah tries to access the ANT+
stick.
It seems something like this has been reported before; I found a
few hits yesterday when I encountered the problem, but didn't see
a solution in those mails.
panic: ehci_device_c
I think this problem is particular to these machine
because I cannot reproduce in my other openbsd
machines when using same kernel and usb dongle...
serial transcript - dmesg + ddb - attached
- cannot paste well because of gmail.
transcript
Description: Binary data
