First off, I spent 1/2 a day searching for an answer to this through forums, FAQs, erratas and just plain google and I could not find an answer. I am however new to the at91 and if this has been covered somewhere I do apologise.

I am using the Adeno Win CE BSP for the AT91SAM9G20-EK and I am having an issue when I try and utilise the board as a Mass Storage Device.

The problem I am seeing is that I am getting an interrupt on Endpoint 0 (UDP_ISR bit 0 is set) signalling that Setup data was received (UDP_CSR0 bit 2 is set) but the length is 0 (UDP_CSR0 bits 16 - 23 are all clear).

The sequence of events that lead up to this is:

• Enumeration takes place and is successful.

• The host issues a 'Read Format Capacities' command to endpoint 2. (This is a Mass Storage command)

• The device acknowledges the command and sets the FORCE STALL bit of both the in and out bulk transfer endpoints (UDP_CSR1 bit 5 and UDP_CSR2 bit 5 are both set) so that the device responds with a stall packet. (Everything to here is as it should be and all is swell)

• The host then issues a 'Clear Feature - Halt Endpoint' request on endpoint 0. (A standard flash drive gets to this point, acknowledges this and continues happily)

• The device receives the interrupt on Endpoint 0 with the rxsetup bit set but data length is 0. (This is the problem!)

• The device stalls Endpoint 0.

The device does eventually recover from this, but it adds a considerable delay to the mounting of the Mass Storage Device.

I can see on my USB protocol analyser (which is a proper hardware based protocol analyser) that the host is sending the correct 8 bytes for the 'Clear Feature - Halt Endpoint' request on endpoint 0 and the device is responding with a stall on Endpoint 0.

If I ignore the length and read out 8 bytes, the data from the UDP_FDR0 register is rubbish.

50 views 0 replies.

Does this mean I am the only one having this sort of difficulty with the USB port or that nobody else has tried to use the USB port in this way?

Perhaps it illustrates how popular WinCE is? It is not clear from the title of the post, or the general forum it is posted to that this is a WinCE question. Or it could be the GoogleBot.

So, have you tried the 9G20-EK usb-device-massstorage-project-at91sam9g20-ek-xxx package? Unless I'm mistaken this should run on a system unencumbered with any OS.

I'am also pretty sure you can contract support directly with Adeno for their BSP, or Atmel FAE's if the aforementioned package doesn't work.

Yes I am getting that impression. This is however not an OS related question but rather one relating to this chip.



Yes, and this package works. I am trying to port it across now as a Win CE driver.

I have taken this up with Atmel and they have been useless. They are quick to answer but constantly ask for more irrelevant data just wasting my time.

The problem to me is clearly a problem with the Atmel processor (trust me, I have ruled all other possibilities out). Under certain circumstances, a 0 length RXSETUP packet can be received. This is rather annoying as the processor has by this stage already acknowledged the message to the host, so the message is lost. My question to Atmel was under what circumstances can this happen and what can be done to avoid it? I would imagine as it is 100% reproducible and because I sent them steps and code this should be rather easy for them to answer but apparently not.

I posted here hoping someone may have ran into this issue and found a solution but that doesn't appear to be the case. If I come to any conclusions I will post them here for future reference.