Why do we try to enter U1 mode in SuperSpeed?

[Hans Petter Selasky]

On 03/30/15 11:19, Hans Petter Selasky wrote:
> On 03/30/15 10:54, Hans Petter Selasky wrote:
>> On 03/30/15 06:54, Kohji Okuno wrote:
>>> Hi HPS
>>>
>>> I have a question.
>>> Why do we try to enter U1 mode in SuperSpeed?
>>> This is the following codes(***)
>>>
>>> In our proprietary environmen(This has a xHCI controller), when a
>>> device accept U1, the device may be strange status. Acutually, when I
>>> tried to connect "Western Digital My Passport Essential 320GB
>>> (P/N:WDBACY3200ABK-00)", this HDD was not recognized correctly.
>>> But, in PC environment, this HDD was recognized correctly.
>>>
>>> Regards,
>>>   Kohji Okuno
>>>
>>> (***)
>>> usb_hub.c:
>>> 793            if (speed == USB_SPEED_SUPER) {
>>> 794                    err = usbd_req_set_hub_u1_timeout(udev, NULL,
>>> 795                        portno, 128 - (2 * udev->depth));
>>> 796                    if (err) {
>>> 797                            DPRINTFN(0, "port %d U1 timeout "
>>> 798                                "failed, error=%s\n",
>>> 799                                portno, usbd_errstr(err));
>>> 800                    }
>>> 801                    err = usbd_req_set_hub_u2_timeout(udev, NULL,
>>> 802                        portno, 128 - (2 * udev->depth));
>>> 803                    if (err) {
>>> 804                            DPRINTFN(0, "port %d U2 timeout "
>>> 805                                "failed, error=%s\n",
>>> 806                                portno, usbd_errstr(err));
>>> 807                    }
>>> 808            }
>>
>> Hi,
>>
>> We are not trying to enter U1 mode, we only setup the U1 and U2
>> timeouts. Is there a flag in the USB descriptors that say this is not
>> supported and that we need to set some other value? I thought all
>> superspeed devices had to support these features and that switching in
>> and out of the sleep modes was done by the XHCI hardware.
>>
>> --HPS
>>
>
> Hi,
>
> Looking at the USB 3.0 specification, I think there might be a typo there:
>
> In table 10-13 - "U2 timeout value encoding", it says that 0xFE
> corresponds to 65024 ms, but I think it should be 65024 _us_ !
>
> Try programming values 0xFF for both U1 and U2. Does the device
> enumerate then?
>
> Maybe it only doesn't support sleep modes during enumeration and that
> after set configuration it works?
>

Hi,

There is a super-speed device capability descriptor. What values does 
the WD HDD say it supports? Currently FreeBSD does not use these values 
and assume that all devices are working below the programmed defaults. 
Hence this descriptor is part of the configuration descriptor, maybe 
sleep modes are not supported until the configuration is set, or maybe 
some firmware devs think so?

--HPS