On Thu, Aug 21, 2014 at 10:00:46AM +0200, Lukasz Majewski wrote:
Hi Felipe,
On Wed, Aug 20, 2014 at 09:34:13AM +0200, Lukasz Majewski wrote:
Hi Felipe,
On Tue, Aug 19, 2014 at 09:08:00PM +0530, Kishon Vijay Abraham I wrote:
On Monday 18 August 2014 08:26 PM, Lukasz Majewski wrote:
Hi Kishon,
>> Explicity set the max packet size in the device descriptor >> to 0x40 as specified in the device class specification for >> device firmware upgrade. Also changed debug to printf to >> explicitly notify the user if the device has been >> enumerated.
It seems like this patch slipped without my attention ...
Hence, I'd like to ask why you must hardcode the packet size here?
http://www.usb.org/developers/devclass_docs/usbdfu10.pdf (in section 4.2.1 DFU Mode Device Descriptor), specifies the bMaxPacketSize0 can be 8,16,32,64. And the value we get from gadget is 512. So hard coded it.
as I mentioned, this should be changed on connection done interrupt.
In the u-boot it is set up during binding the composite.
look at dwc3. dwc3 has a connection done IRQ which also tells me the negotiated link speed.
My point is that you will need to pool in u-boot to emulate this IRQ.
you will need to poll anyways for a bunch of other things. There are some details in DWC3 which are mandated that you wait for event X before doing Y. For example control transfers, it's mandatory that you wait for a XferNotReady event before trying to start a Status Phase.
DFU has one of the poorest documentations for a USB Class and has some moronic definitions.
Frankly speaking I prefer to have the worst standard than some "brilliant" in house developed protocols.
nobody is talking about changing the protocol. DFU was written back when Full speed was the fastest USB on earth.
As you can see the standard is a bit old now... and should be revised.
right, once USB-IF does that, I hope they switch to bulk endpoints :-p