[U-Boot] [PATCH v4 1/2] usb: ehci: add Faraday USB 2.0 EHCI support

[Kuo-Jung Su]

2013/5/8 Marek Vasut <marex at denx.de>:
> Dear Kuo-Jung Su,
>
> [...]
>
>> >> --- a/common/usb_hub.c
>> >> +++ b/common/usb_hub.c
>> >> @@ -419,6 +419,14 @@ static int usb_hub_configure(struct usb_device
>> >> *dev)
>> >>
>> >>                       portstatus = le16_to_cpu(portsts->wPortStatus);
>> >>                       portchange = le16_to_cpu(portsts->wPortChange);
>> >>
>> >> +#ifdef CONFIG_USB_EHCI_FARADAY
>> >> +                     /* Faraday EHCI needs a long long delay here */
>> >> +                     if (!portchange && !portstatus) {
>> >> +                             if (get_timer(start) < 250)
>> >> +                                     continue;
>> >> +                     }
>> >> +#endif
>> >
>> > I'd say just call a weak function here, in case some other non-EHCI
>> > compliant controller happened to need this too. btw. does it need to be
>> > 250 ms or can you poll for readiness somehow ?
>>
>> Got it, thanks. I'll add a weak function later,
>> and about the 250 ms is actually an estimated value.
>> The delay time is actually board specific, it looks to me
>> that it's somehow related to the number of usb host controllers
>> and the attached usb flash drivers.
>>
>> For example:
>>
>> 1.  A369 - FUSBH200: a usb flash driver attached
>>      A369 - FOTG210: nothing attached
>> => no extra delay required.
>>
>> 2.  A369 - FUSBH200: nothing attached
>>      A369 - FOTG210: a usb flash driver attached
>> => no extra delay required.
>>
>> 3. A369 - FUSBH200: a usb flash driver attached
>>     A369 - FOTG210: a usb flash driver attached
>> => The 2nd ehci host requires 200 ms extra delay to detect the attached
>> device. So I put a 250ms here for safe.
>
> Urgh, isn't it a PHY problem then? Or can this not be solved like
> board/genesi/mx51_efikamx/efikamx-usb.c board_ehci_hcd_postinit() or such
> function?
>

It looks to me that it's something related to root hub reset.
And now I've found that adding a extra delay to usb_hub_power_on()
would help to resolve this issue.
So I plan to add a new macro constant for it.
For example:

1. Add this to README:
CONFIG_USB_HUB_MIN_POWER_ON_DELAY defines the minimum
interval for usb hub power-on delay.(minimum 100msec)

2. Update usb_hub.c as bellow:
static void usb_hub_power_on(struct usb_hub_device *hub)
{
......
    /* Wait at least 100 msec for power to become stable */
    mdelay(max(pgood_delay, USB_HUB_MIN_POWER_ON_DELAY));
}

> [...]
>
> Best regards,
> Marek Vasut



--
Best wishes,
Kuo-Jung Su