kern/39727: race condition in ehci isoc abort xfer path

[jmcneill%invisible.ca@localhost]

>Number:         39727
>Category:       kern
>Synopsis:       race condition in ehci isoc abort xfer path
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Oct 11 21:40:00 +0000 2008
>Originator:     Jared D. McNeill
>Release:        NetBSD 4.99.72
>Organization:
        
>Environment:
        
        
System: NetBSD black.invisible.ca 4.99.72 NetBSD 4.99.72 (GENERIC) #173: Sat 
Oct 11 16:43:46 EDT 2008 
jmcneill%black.invisible.ca@localhost:/export/home/jmcneill/branches/HEAD/src/sys/arch/amd64/compile/GENERIC
 amd64
Architecture: x86_64
Machine: amd64
>Description:
        When aborting an ehci isoc transfer using usbd_abort_pipe, a
        panic may happen in usb_transfer_complete.

        ddb appears to skip some frames, but the trace shows:
          usb_transfer_complete
          ehci_softintr

        Workaround: I am unable to reproduce this with a UP kernel.
>How-To-Repeat:
        1. read from a high speed isoc endpoint
        2. create disk activity (start firefox3, thunderbird, vlc in parallel)
        3. once the disk settles, stop reading from the endpoint
        4. panic
>Fix:

>Unformatted: