DFU leave issue on STM32G0B1

[eraquin]

Using STM32G0B1 device, DFU FW update is working as expected
Including leave request, wich results in expected device leave and start at requested address
But `dfu-util` returns error status due to get_status error after leave request

As stated in "AN3156 : USB DFU protocol used in the STM32 bootloader" documentation,
Once file download is complete, host submits leave request
And expects the device to return OK from following get_status
But STM32G0B1 returns `LIBUSB_ERROR_PIPE (-9)` before leaving

Same issue using `STM32CubeProgrammer`, either from CLI or GUI
Need to increase verbosity level for the error log to show up

`dfu-util` and `STM32CubeProgrammer` logs attached

Do I miss something ?
Feel free to ask for any additional information

[STOne-32]

Dear @eraquin ,

The protocol implementation is described here : https://www.st.com/resource/en/application_note/an3156-usb-dfu-protocol-used-in-the-stm32-bootloader-stmicroelectronics.pdf

see Figure 5 ans Figure 6 , Once the DOWNLOAD Command is sent with 0 Data the device trig a reset and enumeration, then the Host is not expected to send a Get_Status as in this case it will be stalled and so as seen in the Logs . May be to see the Manifest bit if set correctly at descriptor level . 

I would not expect any error after this operations as the device already left USB and is jumping to its final application. 

Hope it helps ,

STOne-32

[eraquin]

Thanks a lot for your quick reply

> The protocol implementation is described here
Thanks for confirmation, this is the application note I was looking at

> Once the DOWNLOAD Command is sent with 0 Data the device trig a reset and enumeration,
> then the Host is not expected to send a Get_Status
This is not what I got from the specification
As per the "5.5 Leave DFU mode" section :
> The operation is effectively executed only when a DFU_GETSTATUS request is issued by the host
So from my current understanding, after initiating the leave (download without data),
It is expected for the host to send a get_status status request, to actually trigger the leave
I think I'm able to confirm that this get_status is mandatory
As a host leave request without any following get_status doesn't result in any leave/reset on device side

Still from "5.5 Leave DFU mode" section :
> If the status returned by the device is not dfuMANIFEST, an error has occurred
And here is my concern, as returned status isn't the expected one

[STOne-32]

Dear @eraquin 

Very clear,  let's focus on this at the end indeed.  we will try to see if other MCUs behaving on same way ? or this is particular to this device .  I assume all is running embedded ROM Bootloader ?   as we have the Source sode of DFU implementation and we can run it from user flash as well.  @FBL  your help is appreciated.

Still from "5.5 Leave DFU mode" section :
> If the status returned by the device is not dfuMANIFEST, an error has occurred
And here is my concern, as returned status isn't the expected one

Ciao

STOne-32.

[eraquin]

Thanks for quick reply

> other MCUs behaving on same way ?
I can tell that some STM32 devices are not behaving this way
For instance, no issue with STM32F7x3IxK, which is used on a former product on our side

> I assume all is running embedded ROM Bootloader ?
Yes, DFU running from ROM, on both STM32G0B1 and STM32F7x3IxK