rshartog
commented
8 years ago Hi Eero,
I have never done a pull-request before, so hopefully this is correct. I have done fairly extensive testing of the change on my side -- including a final check of wiping my local stuff and testing straight from the branched code.
If you run the TinyWire_Stress_Master (Uno or other standard Arduino) and TinyWire_Stress_Slave (Tiny85) programs with the current usiTwiSlave.c file, you will see one successful transaction, data mismatches on several subsequent transactions, and then the usiTwiSlave.c will hang on line 406, "while ( tmphead == txTail );". This behavior is consistent with the usiTwiSlave.c issue addressed in this change.
If you run the same master/slave programs with usiTwiSlave.c file I submitted (1 lined moved), it will run successfully all day. I have typed some descriptions of the problem in the master/slave programs themselves.
I also have another, less fundament proposal to allow the full size of the Tx and Rx buffers to be used. In the current implementation in usiTwiSlave.c, the Tx and Rx buffers cannot fully wrap where the head equals the tail, so the available buffer sizes are really TWI_RX_BUFFER_SIZE - 1 and TWI_TX_BUFFER_SIZE - 1. This matters because the master program thinks it can send a specific number of bytes (TWI_RX_BUFFER_SIZE) in a transmit begin()/end() set and this number is really one less than the code would make it seem. The same is true for the requestEvent callback on the slave putting data into the Tx buffer.
Let me know if you're open to this change and I'll do another pull-request. The change would affect more lines (probably about 12) in usiTwiSlave.c, but the change will be obvious and it will not be nearly as subtle as the "single-byte send" issue in usiTwiSlave.c.
Thanks for all your work on TinyWireS, --Scott
rambo
Summary: To allow the master to request more than one byte from the slave, the USI_REQUEST_CALLBACK() call on line 613 of the current usiTwiSlave.c code should be on line 583.
Explanation: For the current code, the ISR routine correctly sends the execution through the “USI_SLAVE_SEND_DATA” case-branch for each byte sent from the slave to the master. With the USI_REQUEST_CALLBACK() on line 613, the callback routine is called for each byte that the master requests. Assuming the callback should queue the send data for the entire request (my assumption), then the current code generates N bytes of data for each byte sent (N^2 bytes total) and only removes N bytes from the transmit buffer. This works when N==1, but if N>1 this quickly results in a transmit buffer overflow and execution hangs on line 406. (This behavior was observed by having the slave respond with the values of txHead and txTail until the slave hung on line 406.)