Re: Oops in UHCI when encountering "host controller process error"

[Jeremy Fitzhardinge <jeremy@goop.org>]

Alan Stern wrote:
>> ------------[ cut here ]------------
>> WARNING: at /home/jeremy/hg/xen/paravirt/linux/drivers/usb/host/uhci-q.c:1744 uhci_scan_schedule+0xb7/0x9a3()
>> uhci->skelqh[1]->node.next is NULL
>>     
>
> Odd.  Is skelqh[1] the only entry whose node.next is NULL?  Not 
> skelqh[2] or others?
>   

No, the others too, but all the WARN_ON backtraces got very verbose and 
I had to rush home at the end of the day.

> The node is initialized during uhci_start() in uhci-hcd.c -- there's a
> loop which initializes all the uhci->skelqh[i] entries by calling
> uhci_alloc_qh().  You should check inside that loop to see whether
> uhci->skelqh[i]->node.next is NULL to begin with.  If it is then
> something about the DMA-coherent memory is funny; if not then the
> linked list pointer gets corrupted somewhere.
>   

OK, will check it out.

    J