Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> writes:
> On Thu, 13 Aug 2009, Vernon Sauder wrote:
>
>> Robert,
>>
>> In our testing, we found that the HW-bit solution for single-threading
>> the setup requests is not sufficient to eliminate the "invalid state"
>> warning message. It seems that the ACM bit is not honored for set
>> configuration packets. We saw the host send an IN packet directly after
>> the set conf before the gadget had time to perform the callback to the
>> udc. We did not see any ill effect on the interaction by hitting this
>> stage.
Let me understand that. The sequence you see is :
Host pxa27x_udc
| |
| ------ ep0(SETUP) ------> |
| ------- ep0(OUT) -------> |
| ------- ep0(OUT) -------> |
| ------- ep0(OUT) -------> |
| SetConfig() |
| |
| ----- ep0(IN)token -----> |
| <---- ep0(IN, NACK)) ---- |
| |
| ----- ep0(IN)token -----> |
| <---- ep0(IN, NACK) ----- |
. .
. . Time passes, and the UDC callback is handled
. .
| ----- ep0(IN)token -----> |
| <----- ep0(IN, ACK) ----- |
| Status stage |
| |
This sequence is the one I was expecting when I wrote the code. As far as my
understanding goes, this is correct, in regard to the USB 2.0 specification
(chapter 8.5.3.1, "function busy").
I don't understand why the warning is raised. Would you be so kind as to send me
the "beagle" capture file, because I'm a bit blind without it. I won't take any
action before I see what's actually happening.
> Even though you didn't see any ill effect, it can nevertheless cause
> problems. If the host starts sending requests before the Set-Config
> has completed, it may end up trying to communicate with endpoints
> before they are enabled. That won't be good.
Agreed.
Cheers.
--
Robert
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
