No subject

delay
is more than 20ms, so caused the bad performance.

Also I am not sure if EHCI can read the old hw_token correctly in this kind=
 of
inconsistent memory state.

>
> Some EHCI implementations have a quirk, in which they perform the
> overlay even when ACTIVE is clear. =A0But even these implementations
> won't try to execute the qTD, because the old value of dummy->hw_token
> also has the HALT bit set.
>
>> It is not only a reasoning or guess, and I have traced this kind of
>> fact certainly.
>
> If your controller behaves as you suggest then it is buggy. =A0And in
> that case, adding another memory barrier won't fix it. =A0There is still
> the possibility that the HC will read dummy during the brief time after
> the existing wmb() and before the CPU has written dummy->hw_token.

No, the mb after 'dummy->hw_token=3Dtoken' does fix the problem. As
said above, IOC IRQ is surely delayed from view of CPU.


thanks,
--
Ming Lei