Jonathan de Boyne Pollard wrote in a message to Mike Bilow:

  > SIO creates alias devices in addition to the standard ones.  For example,
  > COM1 is aliased by $SIO$1, COM2 is aliased by $SIO$2, etc.
  > Since locking and exclusion are handled by the file system
  > router, there is no contention if the device names are
  > different.

 JdBP>   If I recall correctly from when I sat down and
 JdBP> experimented, they   aren't mirror images of the COMn
 JdBP> devices.  Certainly reading data from   and writing data to
 JdBP> the $SIO$n devices produced wildly different   results to
 JdBP> doing the same with the COMn devices.  I suspect that most  
 JdBP> of the API of the $SIO$n devices is via IOCtls.

I've never really looked at it, and SIO.SYS goes to rather elaborate
lengths to prevent disassembly.  For example, it is the only OS/2 device
driver that goes out as an LX-format executable rather than an NE-format
executable, and the code is LZ-compressed.  Speaking as the proud owner of
bug PJ18521, I can assure you that this is far from a good idea.

If I were writing SIO, I would certain make sure that the only interface to
the alias devices was by IOCtl.  You could probably trace what is happening
in SIO with the debugging kernel, but I never bothered.
 
-- Mike


---