PXA: Handling of UDC with no VBUS pins

PXA: Handling of UDC with no VBUS pins

[Dmitry Eremin-Solenikov]

Hello, colleagues,

There are several PXA platforms which on the UDC tranceiver part have
a pullup pin to support connect/disconnect but no VBUS pin to detect
host presense. Currently such platforms are supported through udc driver
itself. I'd like to clean that part of pxa25x_udc/pxa27x_udc drivers and
to move gpio pullup handling to tranceiver.

My question is what would be better: to extend gpio-vbus to work w/o
VBUS pin or to write new transceiver driver only handling gpio-pullup?

-- 
With best wishes
Dmitry

PXA: Handling of UDC with no VBUS pins

[Robert Jarzmik]

On 05/18/2011 02:06 PM, Dmitry Eremin-Solenikov wrote:
> Hello, colleagues,
>
> There are several PXA platforms which on the UDC tranceiver part have
> a pullup pin to support connect/disconnect but no VBUS pin to detect
> host presense. Currently such platforms are supported through udc driver
> itself. I'd like to clean that part of pxa25x_udc/pxa27x_udc drivers and
> to move gpio pullup handling to tranceiver.
So, you want to concentrate USB pullup handling, right ?
> My question is what would be better: to extend gpio-vbus to work w/o
> VBUS pin or to write new transceiver driver only handling gpio-pullup?
My personnal feeling is neither, let them be. The generic pullup would 
have to ask itself :
   - should the pullup be applied on D+ (on USB 2.0 and 3.0 devices) ?
   - should the pullup be applied on D- (on USB 1.1 devices) ?
The common driver will have to handle the suspend/resume in an 
homogenous way.
If the gadget driver is removed, the pullup should be pulled low.

All of this work for only 2 drivers looks overkill to me.

And if you want to gather it in a driver, "gpio-vbus" purpose was to 
handle VBUS sensing, ie. power presence on +5V/Gnd lines. The D+ 
handling was not part of its duties, and you'll probably end up with a 
new driver, which is ... overkill IMHO.

Cheers.

--
Robert

PXA: Handling of UDC with no VBUS pins

[Dmitry Eremin-Solenikov]

On 5/20/11, Robert Jarzmik <robert.jarzmik@free.fr> wrote:
> On 05/18/2011 02:06 PM, Dmitry Eremin-Solenikov wrote:
>> Hello, colleagues,
>>
>> There are several PXA platforms which on the UDC tranceiver part have
>> a pullup pin to support connect/disconnect but no VBUS pin to detect
>> host presense. Currently such platforms are supported through udc driver
>> itself. I'd like to clean that part of pxa25x_udc/pxa27x_udc drivers and
>> to move gpio pullup handling to tranceiver.
> So, you want to concentrate USB pullup handling, right ?

Basically I'd like to cleanup this part of pxa udc drivers.

>> My question is what would be better: to extend gpio-vbus to work w/o
>> VBUS pin or to write new transceiver driver only handling gpio-pullup?
>
> My personnal feeling is neither, let them be. The generic pullup would
> have to ask itself :
>    - should the pullup be applied on D+ (on USB 2.0 and 3.0 devices) ?
>    - should the pullup be applied on D- (on USB 1.1 devices) ?

gpio-vbus driver doesn't do this. It simply handles D+ pullup. I think I'd
start with simple D+ pullup driver and the one may add D- functionality
if required.

> The common driver will have to handle the suspend/resume in an
> homogenous way.

Do pxa25x_udc/pxa27x_udc handle pullup during suspend/resume?

> If the gadget driver is removed, the pullup should be pulled low.
>
> All of this work for only 2 drivers looks overkill to me.

3 boards only from PXA. I don't know about other platforms/soc-families.

> And if you want to gather it in a driver, "gpio-vbus" purpose was to
> handle VBUS sensing, ie. power presence on +5V/Gnd lines. The D+
> handling was not part of its duties, and you'll probably end up with a
> new driver, which is ... overkill IMHO.

gpio-vbus handles D+ already (except suspend/resume).

-- 
With best wishes
Dmitry

PXA: Handling of UDC with no VBUS pins

[Robert Jarzmik]

On 05/21/2011 08:35 PM, Dmitry Eremin-Solenikov wrote:
> My question is what would be better: to extend gpio-vbus to work w/o
>>> VBUS pin or to write new transceiver driver only handling gpio-pullup?
>> My personnal feeling is neither, let them be. The generic pullup would
>> have to ask itself :
>>     - should the pullup be applied on D+ (on USB 2.0 and 3.0 devices) ?
>>     - should the pullup be applied on D- (on USB 1.1 devices) ?
> gpio-vbus driver doesn't do this. It simply handles D+ pullup. I think I'd
> start with simple D+ pullup driver and the one may add D- functionality
> if required.
OK, as long as you don't touch the code in pxa27x_udc.c.
>> The common driver will have to handle the suspend/resume in an
>> homogenous way.
> Do pxa25x_udc/pxa27x_udc handle pullup during suspend/resume?
At least pxa27x_udc does, on suspend to RAM (you should look at 
pxa_udc_suspend()).
>> If the gadget driver is removed, the pullup should be pulled low.
>>
>> All of this work for only 2 drivers looks overkill to me.
> 3 boards only from PXA. I don't know about other platforms/soc-families.

One other point I've forgotten is that pxa27x_udc in its current state 
only supports gpio pullup/pulldown (ie. an external transciever handles 
D+ stimulation). But the PXA sillicon has an internal transciever which 
can be handled directly by pxa27x_udc (register UP2OCR).

Now, nobody uses it. If a board manufacturer is crazy enough to use it, 
the pullup can only be done from inside the pxa27x_udc driver (IO 
mappings driving).


I still think this driver is a bad idea. But if you wish to do it, go 
ahead. The only thing I'll block is removing the existing pullup code 
from pxa27x_udc.c.

Cheers.

--
Robert

PXA: Handling of UDC with no VBUS pins

[Dmitry Eremin-Solenikov]

On Sun, May 22, 2011 at 12:51 PM, Robert Jarzmik <robert.jarzmik@free.fr> wrote:
> On 05/21/2011 08:35 PM, Dmitry Eremin-Solenikov wrote:
>>
>> My question is what would be better: to extend gpio-vbus to work w/o
>>>>
>>>> VBUS pin or to write new transceiver driver only handling gpio-pullup?
>>>
>>> My personnal feeling is neither, let them be. The generic pullup would
>>> have to ask itself :
>>> ? ?- should the pullup be applied on D+ (on USB 2.0 and 3.0 devices) ?
>>> ? ?- should the pullup be applied on D- (on USB 1.1 devices) ?
>>
>> gpio-vbus driver doesn't do this. It simply handles D+ pullup. I think I'd
>> start with simple D+ pullup driver and the one may add D- functionality
>> if required.
>
> OK, as long as you don't touch the code in pxa27x_udc.c.

Let's do it step by step. My plan would be to drop all non-tranceiver code
for handling pullup and VBUS from both pxa25x_udc and pxa27x_udc.

>>>
>>> The common driver will have to handle the suspend/resume in an
>>> homogenous way.
>>
>> Do pxa25x_udc/pxa27x_udc handle pullup during suspend/resume?
>
> At least pxa27x_udc does, on suspend to RAM (you should look at
> pxa_udc_suspend()).

I see... Thanks for pointing to it.

>>>
>>> If the gadget driver is removed, the pullup should be pulled low.
>>>
>>> All of this work for only 2 drivers looks overkill to me.
>>
>> 3 boards only from PXA. I don't know about other platforms/soc-families.
>
> One other point I've forgotten is that pxa27x_udc in its current state only
> supports gpio pullup/pulldown (ie. an external transciever handles D+
> stimulation). But the PXA sillicon has an internal transciever which can be
> handled directly by pxa27x_udc (register UP2OCR).
>
> Now, nobody uses it. If a board manufacturer is crazy enough to use it, the
> pullup can only be done from inside the pxa27x_udc driver (IO mappings
> driving).

Actually I've thought about this. At least several boards use UP2OCR
handling (balloon3, collibri, cm-x300, em-x270, spitz, stargate, voipac, zeus).

> I still think this driver is a bad idea. But if you wish to do it, go ahead.
> The only thing I'll block is removing the existing pullup code from
> pxa27x_udc.c.

Even if I provide full replacement (from the pulldown-on-suspend POV)?

-- 
With best wishes
Dmitry

PXA: Handling of UDC with no VBUS pins

[Sergei Shtylyov]

Hello.

On 20-05-2011 22:12, Robert Jarzmik wrote:

>> There are several PXA platforms which on the UDC tranceiver part have
>> a pullup pin to support connect/disconnect but no VBUS pin to detect
>> host presense. Currently such platforms are supported through udc driver
>> itself. I'd like to clean that part of pxa25x_udc/pxa27x_udc drivers and
>> to move gpio pullup handling to tranceiver.

> So, you want to concentrate USB pullup handling, right ?

>> My question is what would be better: to extend gpio-vbus to work w/o
>> VBUS pin or to write new transceiver driver only handling gpio-pullup?

> My personnal feeling is neither, let them be. The generic pullup would have to
> ask itself :
> - should the pullup be applied on D+ (on USB 2.0 and 3.0 devices) ?
> - should the pullup be applied on D- (on USB 1.1 devices) ?

    Pullup on D- is only used on low speed devices, not all USB 1.1 devices. 
That's how they get detected.

WBR, Sergei