TryReadWrite function

[Kirill Bagrinovsky]

Hello
I read source code of com0com during 3 month. And I have some questions.

1. I see startRead and firstRead flags in TryReadWrite function. Why you use it? firstRead = true when we take first Irp from queue and it not canceled. Is it fact significant for us?

Why we can't write something like this:

>> dataIrpRead.data.irp.pIrp = StartCurrentIrp(pQueueRead, &pCancelRoutineRead, &firstRead);

instead of

>>  if (startRead) {
>>    dataIrpRead.data.irp.pIrp = pQueueRead->pCurrent;
>>    pCancelRoutineRead = NULL;
>>  } else {
>>    dataIrpRead.data.irp.pIrp = StartCurrentIrp(pQueueRead, &pCancelRoutineRead, >>&firstRead);
>>  }

2. Why you use so complex scheme ( TX QUEUE -> TX FIFO -> TX shift register -> Rx shift register -> Rx Fifo -> Rx Buffer -> Rx Queue)? Why not just a Tx Queue -> Rx Queue?

[Vyacheslav Frolov]

1. It's not a trivial task to describe.
2. For more true emulation of real serial port. Old versions had just a Tx Queue -> Rx Queue.

[Kirill Bagrinovsky]

Vyacheslav
Can answer me on some questions about com0com source code

1. Function ReadFromBuffers. In start of this function we call ReadFromBuffer and then ReadFromTxBuffer. Why we pass [&pReadIoPort->readBuf] parameter to ReadFromTxBuffer? I think that readBuf is already empty when we call ReadFromTxBuffer. Isn't it?

2. Function FdoPortQueryInformation.

if (pIrpStack->Parameters.DeviceIoControl.OutputBufferLength < sizeof(FILE_STANDARD_INFORMATION))
        {
           // <<HERE>>
           status = STATUS_BUFFER_TOO_SMALL;
           break;
        }

Should we add line pIrp->IoStatus.Information = 0 on label <<HERE>>?

3. Function FdoPortStartIrp.

  if (pIrpStack->MajorFunction == IRP_MJ_DEVICE_CONTROL && pIrpStack->Parameters.DeviceIoControl.IoControlCode == IOCTL_SERIAL_WAIT_ON_MASK)
    status = NoPending(pIrp, STATUS_INVALID_PARAMETER);

If understand this place right - here you handle situation when you receive second IOCTL_SERIAL_WAIT_ON_MASK request. And you complete this request with status STATUS_INVALID_PARAMETER.

I read MSDN and find this:
>> An IOCTL_SERIAL_SET_WAIT_MASK request is received while a
>> wait-on-mask request is  pending. The driver completes the
>> pending wait-on-mask request with a status of
>> STATUS_SUCCESS and the output wait mask is set to zero.

4. Function FdoPortStartIrp.

if (pIrpStack->MajorFunction == IRP_MJ_FLUSH_BUFFERS)
  status = NoPending(pIrp, STATUS_SUCCESS);

You try to complete Irp that placed in queue. Am i correct?

5. Function TryReadWrite. I understand why you use flag start read. But why you set pCancelRoutineRead = NULL? Is it necessary?

  if (startRead)
  {
    dataIrpRead.data.irp.pIrp = pQueueRead->pCurrent;
    pCancelRoutineRead = NULL;
  }
  else
    dataIrpRead.data.irp.pIrp = StartCurrentIrp(pQueueRead, &pCancelRoutineRead, &firstRead);

[Vyacheslav Frolov]

1. ReadFromTxBuffer do writing to readBuf.
2. Possible it's a bug.
3. MSDN: "A status of STATUS_INVALID_PARAMETER indicates that no wait events are set, or a wait-on-mask request is already pending."
4. Possible it's a bug.
5. It's for compiling:

     com0com\sys\io.c(1391) : error C2220: warning treated as error - no 'object' file generated
     com0com\sys\io.c(1391) : error C4701: local variable 'pCancelRoutineRead' may be used without having been initialized

   and for debuging:

     HALT_UNLESS(pCancelRoutine);

[Kirill Bagrinovsky]

Hello, Vyacheslav
Can you write some words on my questions to help me understand your code in IO.c file, fuction TryReadWrite, section [Move shifted data from TX QUEUE to RX QUEUE while RX QUEUE is not empty]?

===============================================
if (dataIrpWrite.data.irp.status == STATUS_PENDING)
  break;

if (dataIrpWrite.data.irp.status == STATUS_SUCCESS &&
    pIrpStackWrite->MajorFunction == IRP_MJ_DEVICE_CONTROL &&
    pIrpStackWrite->Parameters.DeviceIoControl.IoControlCode == 
    IOCTL_SERIAL_XOFF_COUNTER)
{
        if(startWrite && firstWrite)
          status = STATUS_PENDING;
        else
          StopCurrentIrp(STATUS_PENDING, firstWrite, doneWrite, pIoPortWrite, pQueueWrite, pQueueToComplete);

        if (dataIrpWrite.data.irp.pIrp == pQueueWrite->pCurrent)
        {
          if (doneWrite)
            StartXoffCounter(pIoPortWrite, pQueueToComplete);

          if (dataIrpWrite.data.irp.pIrp == pQueueWrite->pCurrent)
          {
            dataIrpWrite.data.irp.pIrp = NULL;
            break;
          }
        }
}
==========================================

1. I analyse code and see that [dataIrpWrite.data.irp.status] can have only two values STATUS_PENDING and STATUS_SUCCESS. Why you check [dataIrpWrite.data.irp.status == STATUS_SUCCESS] following code?
>> if (dataIrpWrite.data.irp.status == STATUS_PENDING)
>>   break;

2. In what conditions this IF be FALSE?
>> if (dataIrpWrite.data.irp.pIrp == pQueueWrite->pCurrent)
I think pCurrent element always equal to dataIrpWrite.data.irp.pIrp befor read this code.

3. Why IOCTL_SERIAL_XOFF_COUNTER request get STATUS_PENDING status after following checking?
>> if (dataIrpWrite.data.irp.status == STATUS_PENDING)
>>   break;

[Kirill Bagrinovsky]

And one little question

Why you always check
>> if (IrpReadStatus == STATUS_PENDING)
in [Move shifted data from TX QUEUE to RX QUEUE while RX QUEUE is not empty] section?

I think that IRPs with status [not STATUS_PENDING] don't enter in [while (IrpRead)] loop in section
[Move shifted data from TX QUEUE to RX QUEUE while RX QUEUE is not empty].

[Kirill Bagrinovsky]

I understand anser on question (3).
Processing of SERIAL_XOFF_COUNTER  consist of 2 stages