[U-Boot] [PATCH 5/7] usb: Assure Get Descriptor request is in separate microframe

[Marek Vasut <marex@denx.de>]

On 05/04/2016 10:03 AM, Stefan Roese wrote:
> On 03.05.2016 22:51, Marek Vasut wrote:
>> The Kingston DT Ultimate USB 3.0 stick is sensitive to this first
>> Get Descriptor request and if the request is not in a separate
>> microframe, the stick refuses to operate. Add slight delay, which
>> is enough for one microframe to pass on any USB spec revision.
>>
>> Signed-off-by: Marek Vasut <marex@denx.de>
>> Cc: Chin Liang See <clsee@altera.com>
>> Cc: Dinh Nguyen <dinguyen@opensource.altera.com>
>> Cc: Hans de Goede <hdegoede@redhat.com>
>> Cc: Stefan Roese <sr@denx.de>
>> Cc: Stephen Warren <swarren@nvidia.com>
>> ---
>>   common/usb.c | 8 ++++++++
>>   1 file changed, 8 insertions(+)
>>
>> diff --git a/common/usb.c b/common/usb.c
>> index 205041b..8d9efe5 100644
>> --- a/common/usb.c
>> +++ b/common/usb.c
>> @@ -1077,6 +1077,14 @@ int usb_select_config(struct usb_device *dev)
>>       le16_to_cpus(&dev->descriptor.idProduct);
>>       le16_to_cpus(&dev->descriptor.bcdDevice);
>>
>> +    /*
>> +     * Kingston DT Ultimate 32GB USB 3.0 seems to be extremely sensitive
>> +     * about this first Get Descriptor request. If there are any other
>> +     * requests in the first microframe, the stick crashes. Wait about
>> +     * one microframe duration here (1mS for USB 1.x , 125uS for USB
>> 2.0).
>> +     */
>> +    mdelay(1);
>> +
>>       /* only support for one config for now */
>>       err = usb_get_configuration_len(dev, 0);
>>       if (err >= 0) {
>>
> 
> Again my question, if this problem also occurs on other platforms
> with this USB key. A 1ms delay is not really a big deal, but its
> my general feeling that we should manifest such changes by testing
> on different platforms.

I tested it by connecting the bus analyzer between the stick and socfpga
and between the stick and x86 host. I wouldn't be able to
come up with this solution. btw. any ehci ends up inserting these
control requests into separate microframes, but we need the mdelay
to also cater for ohci/uhci, which has longer frames (1ms).

-- 
Best regards,
Marek Vasut