Context Navigation

Changes between Initial Version and Version 1 of Ticket #12117, comment 10

initial	v1
3	3	It now seems that the problem arises when two ohci interrupt requests "pile up". Specifically, the logs seem to show a failure with a WDH (Writeback Done Head) interrupt being serviced, then an SOF (Start of Frame) interrupt ocurring before the first irq service completes. The SOF interrupt never gets handled, which leaves the Endpoint Descriptor in an "UNLINK" state and stalls the port.
4	4
5		I don't know exactly how overlapping interrupts are supposed to be handled; I assume the later one should be queued up and processed after the first interrupt returns, but I don't know how the details work. With multiple cores is it possible that the two interrupts could be executing simultaneously??
	5	I don't know exactly how overlapping interrupts are supposed to be handled; I assume the later one should be queued up and processed after the first interrupt returns, but I don't know how the details work. With multiple cores is it possible that the two interrupts could be executing simultaneously on two different cores?
6	6
7	7	I wonder if this is a defect in the guest ohci driver, that is stressed by the timing conditions generated by the virtual DevOHCI emulation. I should probably try the latest Linux kernel; the guest is currently