[U-Boot] U-Boot Rockchip dwc_hcd question for low-speed devices

[U-Boot] U-Boot Rockchip dwc_hcd question for low-speed devices

[Josh Sanford]

All,

I am trying to use USB within U-Boot on a Firefly RK3288 (Rockchip) board,
and I am able to enumerate high-speed devices, but not full-speed or
low-speed devices.  I know that the devices are functional, because they
work on the same physical hardware when I boot into Linux.  Note that I'm
using "U-Boot 2014.10-RK3288-02".

As an aside, the built-in "USB 2.0 Hub [MTT]" hub never generates a
USB_PORT_STAT_C_CONNECTION for me, even though USB_PORT_STAT_CONNECTION
goes high immediately after powering on, so I forcibly assert
USB_PORT_STAT_C_CONNECTION to allow the rest of the tree to enumerate.

High-speed devices can be nested as deeply as I wish and still show up
properly, but I can't get low-speed devices (i.e. a keyboard) to work in
any hub port, including any of the ports directly on the board.  All
attempts to read from my low-speed devices return all zeros.  When I dig a
little deeper, I find myself in int dwc_wait_for_complete(struct usb_device
*dev, uint32_t channel_num, uint32_t *hcStat, uint32_t *errCnt) {}.

On the second iteration through the goto loop, hcintn.b.chhltd and
hcintn.b.xacterr are both asserted, but hcintn.b.ack, hcintn.b.nak, and
hcintn.b.nyet are not, and errCnt is only 1, so the code sets *hcStat =
HCSTAT_REINIT and returns.

I do notice that dwc_hcd.c seems not to have any code to support split
transactions (or perhaps I simply haven't seen it yet).  If true, I wonder
whether this code is actually expected to support full-speed and low-speed
devices.  Do you have any thoughts on the matter?  I wish I were more of a
USB expert, but I'm regrettably quite a novice.

Regards,

Josh

[U-Boot] U-Boot Rockchip dwc_hcd question for low-speed devices

[Simon Glass]

+Marek, Stephen

Hi Josh,

On 18 December 2015 at 06:26, Josh Sanford <joshs6502@gmail.com> wrote:
> All,
>
> I am trying to use USB within U-Boot on a Firefly RK3288 (Rockchip) board,
> and I am able to enumerate high-speed devices, but not full-speed or
> low-speed devices.  I know that the devices are functional, because they
> work on the same physical hardware when I boot into Linux.  Note that I'm
> using "U-Boot 2014.10-RK3288-02".
>
> As an aside, the built-in "USB 2.0 Hub [MTT]" hub never generates a
> USB_PORT_STAT_C_CONNECTION for me, even though USB_PORT_STAT_CONNECTION goes
> high immediately after powering on, so I forcibly assert
> USB_PORT_STAT_C_CONNECTION to allow the rest of the tree to enumerate.
>
> High-speed devices can be nested as deeply as I wish and still show up
> properly, but I can't get low-speed devices (i.e. a keyboard) to work in any
> hub port, including any of the ports directly on the board.  All attempts to
> read from my low-speed devices return all zeros.  When I dig a little
> deeper, I find myself in int dwc_wait_for_complete(struct usb_device *dev,
> uint32_t channel_num, uint32_t *hcStat, uint32_t *errCnt) {}.
>
> On the second iteration through the goto loop, hcintn.b.chhltd and
> hcintn.b.xacterr are both asserted, but hcintn.b.ack, hcintn.b.nak, and
> hcintn.b.nyet are not, and errCnt is only 1, so the code sets *hcStat =
> HCSTAT_REINIT and returns.
>
> I do notice that dwc_hcd.c seems not to have any code to support split
> transactions (or perhaps I simply haven't seen it yet).  If true, I wonder
> whether this code is actually expected to support full-speed and low-speed
> devices.  Do you have any thoughts on the matter?  I wish I were more of a
> USB expert, but I'm regrettably quite a novice.

I believe the split transaction issue is currently being resolved for
DWC. See for example Stefan Bruns' series from 13th December. It might
be worth trying that out.

Regards,
Simon

[U-Boot] U-Boot Rockchip dwc_hcd question for low-speed devices

[Josh Sanford]

Thanks Simon.  I'm having difficulty finding anything by Stefan Bruns
around that time frame, or much of anything recent pertaining to DWC.  (I
only checked the u-boot-denx.git and u-boot-rockchip.git branches.)  Am I
looking in the wrong place?

On Sat, Dec 19, 2015 at 3:30 PM, Simon Glass <sjg@chromium.org> wrote:

> +Marek, Stephen
>
> Hi Josh,
>
> On 18 December 2015 at 06:26, Josh Sanford <joshs6502@gmail.com> wrote:
> > All,
> >
> > I am trying to use USB within U-Boot on a Firefly RK3288 (Rockchip)
> board,
> > and I am able to enumerate high-speed devices, but not full-speed or
> > low-speed devices.  I know that the devices are functional, because they
> > work on the same physical hardware when I boot into Linux.  Note that I'm
> > using "U-Boot 2014.10-RK3288-02".
> >
> > As an aside, the built-in "USB 2.0 Hub [MTT]" hub never generates a
> > USB_PORT_STAT_C_CONNECTION for me, even though USB_PORT_STAT_CONNECTION
> goes
> > high immediately after powering on, so I forcibly assert
> > USB_PORT_STAT_C_CONNECTION to allow the rest of the tree to enumerate.
> >
> > High-speed devices can be nested as deeply as I wish and still show up
> > properly, but I can't get low-speed devices (i.e. a keyboard) to work in
> any
> > hub port, including any of the ports directly on the board.  All
> attempts to
> > read from my low-speed devices return all zeros.  When I dig a little
> > deeper, I find myself in int dwc_wait_for_complete(struct usb_device
> *dev,
> > uint32_t channel_num, uint32_t *hcStat, uint32_t *errCnt) {}.
> >
> > On the second iteration through the goto loop, hcintn.b.chhltd and
> > hcintn.b.xacterr are both asserted, but hcintn.b.ack, hcintn.b.nak, and
> > hcintn.b.nyet are not, and errCnt is only 1, so the code sets *hcStat =
> > HCSTAT_REINIT and returns.
> >
> > I do notice that dwc_hcd.c seems not to have any code to support split
> > transactions (or perhaps I simply haven't seen it yet).  If true, I
> wonder
> > whether this code is actually expected to support full-speed and
> low-speed
> > devices.  Do you have any thoughts on the matter?  I wish I were more of
> a
> > USB expert, but I'm regrettably quite a novice.
>
> I believe the split transaction issue is currently being resolved for
> DWC. See for example Stefan Bruns' series from 13th December. It might
> be worth trying that out.
>
> Regards,
> Simon
>

[U-Boot] U-Boot Rockchip dwc_hcd question for low-speed devices

[Simon Glass]

Hi Josh,

On 21 December 2015 at 07:11, Josh Sanford <joshs6502@gmail.com> wrote:
> Thanks Simon.  I'm having difficulty finding anything by Stefan Bruns around
> that time frame, or much of anything recent pertaining to DWC.  (I only
> checked the u-boot-denx.git and u-boot-rockchip.git branches.)  Am I looking
> in the wrong place?

This is just patches at present - see here:

http://patchwork.ozlabs.org/project/uboot/list/?submitter=67055&state=*

Regards,
Simon

>
> On Sat, Dec 19, 2015 at 3:30 PM, Simon Glass <sjg@chromium.org> wrote:
>>
>> +Marek, Stephen
>>
>> Hi Josh,
>>
>> On 18 December 2015 at 06:26, Josh Sanford <joshs6502@gmail.com> wrote:
>> > All,
>> >
>> > I am trying to use USB within U-Boot on a Firefly RK3288 (Rockchip)
>> > board,
>> > and I am able to enumerate high-speed devices, but not full-speed or
>> > low-speed devices.  I know that the devices are functional, because they
>> > work on the same physical hardware when I boot into Linux.  Note that
>> > I'm
>> > using "U-Boot 2014.10-RK3288-02".
>> >
>> > As an aside, the built-in "USB 2.0 Hub [MTT]" hub never generates a
>> > USB_PORT_STAT_C_CONNECTION for me, even though USB_PORT_STAT_CONNECTION
>> > goes
>> > high immediately after powering on, so I forcibly assert
>> > USB_PORT_STAT_C_CONNECTION to allow the rest of the tree to enumerate.
>> >
>> > High-speed devices can be nested as deeply as I wish and still show up
>> > properly, but I can't get low-speed devices (i.e. a keyboard) to work in
>> > any
>> > hub port, including any of the ports directly on the board.  All
>> > attempts to
>> > read from my low-speed devices return all zeros.  When I dig a little
>> > deeper, I find myself in int dwc_wait_for_complete(struct usb_device
>> > *dev,
>> > uint32_t channel_num, uint32_t *hcStat, uint32_t *errCnt) {}.
>> >
>> > On the second iteration through the goto loop, hcintn.b.chhltd and
>> > hcintn.b.xacterr are both asserted, but hcintn.b.ack, hcintn.b.nak, and
>> > hcintn.b.nyet are not, and errCnt is only 1, so the code sets *hcStat =
>> > HCSTAT_REINIT and returns.
>> >
>> > I do notice that dwc_hcd.c seems not to have any code to support split
>> > transactions (or perhaps I simply haven't seen it yet).  If true, I
>> > wonder
>> > whether this code is actually expected to support full-speed and
>> > low-speed
>> > devices.  Do you have any thoughts on the matter?  I wish I were more of
>> > a
>> > USB expert, but I'm regrettably quite a novice.
>>
>> I believe the split transaction issue is currently being resolved for
>> DWC. See for example Stefan Bruns' series from 13th December. It might
>> be worth trying that out.
>>
>> Regards,
>> Simon
>
>

[U-Boot] U-Boot Rockchip dwc_hcd question for low-speed devices

[Josh Sanford]

Thanks Simon,

I've been looking through the code you referenced.  So these patches are
just standalone snippets and not yet part of any branch?  (I'm just not
familiar with U-Boot's development process.)

Josh

On Tue, Jan 5, 2016 at 7:25 PM, Simon Glass <sjg@chromium.org> wrote:

> Hi Josh,
>
> On 21 December 2015 at 07:11, Josh Sanford <joshs6502@gmail.com> wrote:
> > Thanks Simon.  I'm having difficulty finding anything by Stefan Bruns
> around
> > that time frame, or much of anything recent pertaining to DWC.  (I only
> > checked the u-boot-denx.git and u-boot-rockchip.git branches.)  Am I
> looking
> > in the wrong place?
>
> This is just patches at present - see here:
>
> http://patchwork.ozlabs.org/project/uboot/list/?submitter=67055&state=*
>
> Regards,
> Simon
>
> >
> > On Sat, Dec 19, 2015 at 3:30 PM, Simon Glass <sjg@chromium.org> wrote:
> >>
> >> +Marek, Stephen
> >>
> >> Hi Josh,
> >>
> >> On 18 December 2015 at 06:26, Josh Sanford <joshs6502@gmail.com> wrote:
> >> > All,
> >> >
> >> > I am trying to use USB within U-Boot on a Firefly RK3288 (Rockchip)
> >> > board,
> >> > and I am able to enumerate high-speed devices, but not full-speed or
> >> > low-speed devices.  I know that the devices are functional, because
> they
> >> > work on the same physical hardware when I boot into Linux.  Note that
> >> > I'm
> >> > using "U-Boot 2014.10-RK3288-02".
> >> >
> >> > As an aside, the built-in "USB 2.0 Hub [MTT]" hub never generates a
> >> > USB_PORT_STAT_C_CONNECTION for me, even though
> USB_PORT_STAT_CONNECTION
> >> > goes
> >> > high immediately after powering on, so I forcibly assert
> >> > USB_PORT_STAT_C_CONNECTION to allow the rest of the tree to enumerate.
> >> >
> >> > High-speed devices can be nested as deeply as I wish and still show up
> >> > properly, but I can't get low-speed devices (i.e. a keyboard) to work
> in
> >> > any
> >> > hub port, including any of the ports directly on the board.  All
> >> > attempts to
> >> > read from my low-speed devices return all zeros.  When I dig a little
> >> > deeper, I find myself in int dwc_wait_for_complete(struct usb_device
> >> > *dev,
> >> > uint32_t channel_num, uint32_t *hcStat, uint32_t *errCnt) {}.
> >> >
> >> > On the second iteration through the goto loop, hcintn.b.chhltd and
> >> > hcintn.b.xacterr are both asserted, but hcintn.b.ack, hcintn.b.nak,
> and
> >> > hcintn.b.nyet are not, and errCnt is only 1, so the code sets *hcStat
> =
> >> > HCSTAT_REINIT and returns.
> >> >
> >> > I do notice that dwc_hcd.c seems not to have any code to support split
> >> > transactions (or perhaps I simply haven't seen it yet).  If true, I
> >> > wonder
> >> > whether this code is actually expected to support full-speed and
> >> > low-speed
> >> > devices.  Do you have any thoughts on the matter?  I wish I were more
> of
> >> > a
> >> > USB expert, but I'm regrettably quite a novice.
> >>
> >> I believe the split transaction issue is currently being resolved for
> >> DWC. See for example Stefan Bruns' series from 13th December. It might
> >> be worth trying that out.
> >>
> >> Regards,
> >> Simon
> >
> >
>

[U-Boot] U-Boot Rockchip dwc_hcd question for low-speed devices

[Simon Glass]

Hi Josh,

On 8 January 2016 at 06:27, Josh Sanford <joshs6502@gmail.com> wrote:
> Thanks Simon,
>
> I've been looking through the code you referenced.  So these patches are
> just standalone snippets and not yet part of any branch?  (I'm just not
> familiar with U-Boot's development process.)

Please put your reply at the bottom or the threads will get very confusing!

There is some information here that might help:

http://www.denx.de/wiki/U-Boot/Patches

In short, yes these are patches, and will be applied at some point, at
which point they become part of 'mainline' and end up in the next
release. If you test them you can reply with your Tested-by tag which
helps the process.

Regards,
Simon

>
> Josh
>
> On Tue, Jan 5, 2016 at 7:25 PM, Simon Glass <sjg@chromium.org> wrote:
>>
>> Hi Josh,
>>
>> On 21 December 2015 at 07:11, Josh Sanford <joshs6502@gmail.com> wrote:
>> > Thanks Simon.  I'm having difficulty finding anything by Stefan Bruns
>> > around
>> > that time frame, or much of anything recent pertaining to DWC.  (I only
>> > checked the u-boot-denx.git and u-boot-rockchip.git branches.)  Am I
>> > looking
>> > in the wrong place?
>>
>> This is just patches at present - see here:
>>
>> http://patchwork.ozlabs.org/project/uboot/list/?submitter=67055&state=*
>>
>> Regards,
>> Simon
>>
>> >
>> > On Sat, Dec 19, 2015 at 3:30 PM, Simon Glass <sjg@chromium.org> wrote:
>> >>
>> >> +Marek, Stephen
>> >>
>> >> Hi Josh,
>> >>
>> >> On 18 December 2015 at 06:26, Josh Sanford <joshs6502@gmail.com> wrote:
>> >> > All,
>> >> >
>> >> > I am trying to use USB within U-Boot on a Firefly RK3288 (Rockchip)
>> >> > board,
>> >> > and I am able to enumerate high-speed devices, but not full-speed or
>> >> > low-speed devices.  I know that the devices are functional, because
>> >> > they
>> >> > work on the same physical hardware when I boot into Linux.  Note that
>> >> > I'm
>> >> > using "U-Boot 2014.10-RK3288-02".
>> >> >
>> >> > As an aside, the built-in "USB 2.0 Hub [MTT]" hub never generates a
>> >> > USB_PORT_STAT_C_CONNECTION for me, even though
>> >> > USB_PORT_STAT_CONNECTION
>> >> > goes
>> >> > high immediately after powering on, so I forcibly assert
>> >> > USB_PORT_STAT_C_CONNECTION to allow the rest of the tree to
>> >> > enumerate.
>> >> >
>> >> > High-speed devices can be nested as deeply as I wish and still show
>> >> > up
>> >> > properly, but I can't get low-speed devices (i.e. a keyboard) to work
>> >> > in
>> >> > any
>> >> > hub port, including any of the ports directly on the board.  All
>> >> > attempts to
>> >> > read from my low-speed devices return all zeros.  When I dig a little
>> >> > deeper, I find myself in int dwc_wait_for_complete(struct usb_device
>> >> > *dev,
>> >> > uint32_t channel_num, uint32_t *hcStat, uint32_t *errCnt) {}.
>> >> >
>> >> > On the second iteration through the goto loop, hcintn.b.chhltd and
>> >> > hcintn.b.xacterr are both asserted, but hcintn.b.ack, hcintn.b.nak,
>> >> > and
>> >> > hcintn.b.nyet are not, and errCnt is only 1, so the code sets *hcStat
>> >> > =
>> >> > HCSTAT_REINIT and returns.
>> >> >
>> >> > I do notice that dwc_hcd.c seems not to have any code to support
>> >> > split
>> >> > transactions (or perhaps I simply haven't seen it yet).  If true, I
>> >> > wonder
>> >> > whether this code is actually expected to support full-speed and
>> >> > low-speed
>> >> > devices.  Do you have any thoughts on the matter?  I wish I were more
>> >> > of
>> >> > a
>> >> > USB expert, but I'm regrettably quite a novice.
>> >>
>> >> I believe the split transaction issue is currently being resolved for
>> >> DWC. See for example Stefan Bruns' series from 13th December. It might
>> >> be worth trying that out.
>> >>
>> >> Regards,
>> >> Simon
>> >
>> >
>
>

[U-Boot] U-Boot Rockchip dwc_hcd question for low-speed devices

[Stefan Bruens]

On Montag, 11. Januar 2016 09:59:09 CET Simon Glass wrote:
> Hi Josh,
> 
> On 8 January 2016 at 06:27, Josh Sanford <joshs6502@gmail.com> wrote:
> > Thanks Simon,
> > 
> > I've been looking through the code you referenced.  So these patches are
> > just standalone snippets and not yet part of any branch?  (I'm just not
> > familiar with U-Boot's development process.)
> 
> Please put your reply at the bottom or the threads will get very confusing!
> 
> There is some information here that might help:
> 
> http://www.denx.de/wiki/U-Boot/Patches
> 
> In short, yes these are patches, and will be applied at some point, at
> which point they become part of 'mainline' and end up in the next
> release. If you test them you can reply with your Tested-by tag which
> helps the process.

Just a small followup, I have reworked the patches. I have followed Simons? 
suggestion and split the "core" transfer from the chunk_msg loop into a 
seperate function. The SPLIT changes are now much more self contained.

The patches need some cleanup, especially the INTERRUPT transfer support (yay, 
that did not work before), so I will post it soon, but maybe not earlier than 
during the weekend.

Kind regards,

Stefan

-- 
Stefan Br?ns  /  Bergstra?e 21  /  52062 Aachen
home: +49 241 53809034     mobile: +49 151 50412019
work: +49 2405 49936-424