Random data corruption in outTransfer

[randomricky2018]

Hi,

I'm a super beginner in Arduino, usb and these drivers stuff, not sure if I understand the problem correctly. :(

I was trying to write a Android Adb host using this library, and I found that the messages that I sent sometimes (or always) corrupted. (The bytes in Usb.outTransfer does not match what's being received).

Here's my thought: (correct me if i'm wrong :) )

In UsbTransfer, due to nak bug, if nak limit is > 1, it will retry when it receives nak.
To restore it from nak, we do "regWr(rSNDFIFO, *data_p);".
However, *data_p is corrupted (the content is not the original content anymore), as in bytesWr(), we do "USB_SPI.transfer(data_p, nbytes);", which read AND write bytes in data_p by SPI.
Hence, "regWr(rSNDFIFO, *data_p);" does not write the original first byte anymore, it writes a value read from SPI.

Does it make sense or if I misunderstood sth?

Thanks a lot!

[gdsports]

I have seen this problem sending to a slow printer which generates lots of NAKs. The root cause is the SPI.transfer function reads AND writes the data buffer so in Usb.cpp|USB::OutTransfer the following line sends garbage instead of the correct value.

USB_Host_Shield_2.0/Usb.cpp

Line 376 in 49421b6

regWr(rSNDFIFO, *data_p);

bytesWr() calls SPI.transfer() which sends data to the USB out FIFO but also reads from the SPI bus thus over writing the original values at data_p.

USB_Host_Shield_2.0/Usb.cpp

Line 337 in 49421b6

bytesWr(rSNDFIFO, bytes_tosend, data_p); //filling output FIFO

My fix is to save the first byte of data_p before bytesWr() then use it for NAK retry, if needed.

		uint8_t save_byte = *data_p;
                bytesWr(rSNDFIFO, bytes_tosend, data_p); //filling output FIFO


                regWr(rSNDFIFO, save_byte);

#238 might also be related.

[Lauszus]

Fixed in #473