[U-Boot] [linux-sunxi] [PATCH 0/6] sunxi: H6: Enable USB (2.0) support

Thu Jun 20 21:17:30 UTC 2019

On 20/06/2019 19:01, Clément Péron wrote:

Hi Clément,

thanks for trying that!
...

> On Wed, 19 Jun 2019 at 12:03, Andre Przywara <andre.przywara at arm.com> wrote:
>>
>> On Wed, 19 Jun 2019 10:57:14 +0200
>> Clément Péron <peron.clem at gmail.com> wrote:
>>
>> Hi,
>>
>>> On Wed, 19 Jun 2019 at 03:04, André Przywara <andre.przywara at arm.com> wrote:
>>>>
>>>> On Tue, 18 Jun 2019 19:22:34 +0200
>>>> Clément Péron <peron.clem at gmail.com> wrote:
>>>>
>>>> Hi,
>>>>
>>>>
>>>>> On Tue, 18 Jun 2019 at 19:08, Clément Péron <peron.clem at gmail.com> wrote:
>>>>>>
>>>>>> Hi Andre,
>>>>>>
>>>>>> On Tue, 18 Jun 2019 at 09:50, Jagan Teki <jagan at amarulasolutions.com> wrote:
>>>>>>>
>>>>>>> On Thu, May 16, 2019 at 6:56 AM Andre Przywara <andre.przywara at arm.com> wrote:
>>>>>>>>
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> this series enables USB support on the H6 boards. This is mostly just
>>>>>>>> adding some missing pieces here and there, the actual controller and PHY
>>>>>>>> are very similar to the previous ones, if not identical.
>>>>>>>> This is for the 2.0 ports only at the moment, USB 3.0 will be done
>>>>>>>> later (started porting Icenowy's Linux driver).
>>>>>>>> The Pine H64 shares a similar problem as the Pine64+ boards regarding
>>>>>>>> the upper USB port. To enable this port, we need the first patch
>>>>>>>> from the series [1] fixing this issue on the A64 boards.
>>>>>>>>
>>>>>>>> Patch 1 is a drive-by patch to bring SUNXI_GPIO to Kconfig, as this was
>>>>>>>> lingering in one of my branches for a while.
>>>>>>>> Patch 2 enables GPIO support for the H6, as this is needed for the Pine
>>>>>>>> H64 to enable the VBUS regulator.
>>>>>>>> Patch 3 adds the clock and reset gates mappings for the USB controller and
>>>>>>>> the PHY, the values are taken from the manual and verified against
>>>>>>>> Linux.
>>>>>>>> Patch 4 adds some code to the PHY driver to skip over not implemented
>>>>>>>> PHYs, as the H6 uses a PHY0/PHY3 combination in the DT.
>>>>>>>> Patch 5 then eventually enables USB in the existing defconfigs.
>>>>>>>> Patch 6 adds the .dts fixes required to get the upper USB port to work
>>>>>>>> on the Pine H64.
>>>>>>>>
>>>>>>>> Clément, can you please verify that this works for the Beelink box?
>>>>>> Thank you for the notice.
>>>>>>
>>>>>> Works fine on my board:
>>>>> I have tested with and without the trick in usb2otg node and both seems working.
>>>>> (only tested usb start; usb tree).
>>>>
>>>> You mean dropping the status="okay"?
>>> I mean, I have tested with and without the whole patch and it works in
>>> both cases.
>>
>> Do you mean it works without adding "phys = <&usb2phy 0>;" to the ehci0
>> and ohci0 nodes?
>> Are you sure that USB *really* works? The boot log *always* enumerates the
>> bus and prints one device (the root hub), but does "usb tree" show an
>> actual connected device? The easiest test is to connect a pen drive, at
>> which point U-Boot automatically lists the USB mass storage device in the
>> boot log.
> 
> You're right, only the root hub is enumerated.
> Actually the VBUS supply on my board is directly provided by the
> always-on "DC Jack".
> So normally this hack is not needed and the USB should work in both
> case which is kind of expected.

Well, the power part is one thing, but quite some boards do it like this
(Pine64, for instance).
This PHY connection property is something separate, it's just needed for
U-Boot, because the driver will bail out if it cannot find a PHY. It's a
driver shortcoming, if you like, checkout ehci_setup_phy() in
drivers/usb/host/ehci-hcd.c and it's call site in
drivers/usb/host/ehci-generic.c:ehci_usb_probe().

> But unfortunately it's doesn't seem to work. I have tested with a
> storage usb and it's not listed :(

Mmmh, interesting. Does it work in Linux?
Also, did you ever try FEL booting, using this port and an A-to-A cable?
Just asking to make sure that USB 2.0 port is actually USB0, and not the
other USB 2.0 port on the SoC (USB3).

> The TX UART of my board is not functional so it's a bit touchy to
> debug the issue.

Ah, same as on the Eachlink H6 Mini, then. Did you see what Sven and I
came up with?
http://linux-sunxi.org/Eachlink_H6_Mini#Adding_a_serial_port_.28might_void_warranty.3F.29

> Anyway it doesn't make the board reset so it's not blocking.
> I will try to investigate this later.

OK, thanks!

Cheers,
Andre.

>>>> Yeah, I figured as well that it works both ways. Actually older kernels
>>>> require the MUSB driver to be enabled, while newer kernels don't seem to
>>>> care. As this affects the Debian Buster kernel (4.19, in the installer),
>>>> I plan on just leaving it as it is right now.
>>>>
>>>>> Why not leave it to avoid a difference with linux?
>>>>
>>>> Technically we should disable it, as it's not usable and might clash
>>>> with EHCI0/OHCI0, but for said reason I just keep it.
>>> That's what I have in mind, if it's not mandatory just leave it as
>>> this, this will avoid differences with linux.
>>
>> There shall be no real differences with Linux, in this case I will push
>> those changes to Linux, so we sync this somewhat the other way.
>> The hardware clearly connects the USB0 controller to PHY0, the DT should
>> reflect this.
>>
>> Cheers,
>> Andre.
>>
>>>>>> U-Boot 2019.07-rc4 (Jun 18 2019 - 19:06:21 +0200) Allwinner Technology
>>>>>>
>>>>>> CPU:   Allwinner H6 (SUN50I)
>>>>>> Model: Beelink GS1
>>>>>> DRAM:  2 GiB
>>>>>> MMC:   mmc at 4020000: 0, mmc at 4022000: 1
>>>>>> Loading Environment from FAT... Unable to use mmc 1:1... In:    serial at 5000000
>>>>>> Out:   serial at 5000000
>>>>>> Err:   serial at 5000000
>>>>>> Net:   No ethernet found.
>>>>>> starting USB...
>>>>>> Bus usb at 5101000: USB EHCI 1.00
>>>>>> Bus usb at 5101400: USB OHCI 1.0
>>>>>> scanning bus usb at 5101000 for devices... 1 USB Device(s) found
>>>>>>
>>>>>> Regards,
>>>>>> Clément
>>>>>>
>>>>>>>> I guess the USB 2.0 port is probably the OTG one, so this setup would
>>>>>>>> look somewhat similar to the PineH64, which would allow you to copy
>>>>>>>> the USB DT nodes from there.
>>>>>>>>
>>>>>>>> Cheers,
>>>>>>>> Andre.
>>>>>>>>
>>>>>>>> [1] https://lists.denx.de/pipermail/u-boot/2019-May/369520.html
>>>>>>>>
>>>>>>>> Andre Przywara (6):
>>>>>>>>   sunxi: move SUNXI_GPIO to Kconfig
>>>>>>>>   sunxi: gpio: Enable support for H6 pin controller
>>>>>>>>   sunxi: clocks: Add H6 USB clock gates and resets
>>>>>>>>   sunxi: phy: Add USB PHY support for Allwinner H6
>>>>>>>>   sunxi: H6: Enable USB for existing boards
>>>>>>>>   sunxi: Pine64: DTS: enable USB PHY 0 for HCI0
>>>>>>>
>>>>>>> Except 6/6 (it can be part of DTS sync during MW) and rest
>>>>>>>
>>>>>>> Reviewed-by: Jagan Teki <jagan at amarulasolutions.com>
>>>>
>>
>> --
>> You received this message because you are subscribed to the Google Groups "linux-sunxi" group.
>> To unsubscribe from this group and stop receiving emails from it, send an email to linux-sunxi+unsubscribe at googlegroups.com.
>> To view this discussion on the web, visit https://groups.google.com/d/msgid/linux-sunxi/20190619110302.7c4ea7e8%40donnerap.cambridge.arm.com.
>> For more options, visit https://groups.google.com/d/optout.