Setting termios VMIN > 0 and VTIME > 0 on non blocking file

[Corinna Vinschen]

On Mar 12 15:20, Åke Rehnman via Cygwin wrote:
> 
> On 2020-03-12 12:40, Corinna Vinschen wrote:
> > 
> > I inspected the serial I/O read function and I only see a subtil
> > difference in terms of VMIN/VTIME which doesn't seem to be the culprit
> > at first glance.  In O_NONBLOCK mode, the underlying Windows function
> > ReadFile is called unconditionally.  My current hunch is this:
> > 
> > - If VMIN>0 && VTIME>0, the Windows equivalent of tcsetattr is
> >    told that VMIN>0.
> > 
> > - So, assuming VMIN == 2 in Windows.  If VTIME > 0, the Cygwin read
> >    function sets the number of bytes_to_read to 1 (this is old code,
> >    don't ask why).
> Do you have a pointer to this? I can not find it...

That's because it's not true.  I created this mail during lunch break
and screwed up.

> > - My assumption now is that the ReadFile function fails because it's
> >    supposed to return only 1 char, but VMIN is > 1.
> > 
> > There are two issues here.
> > 
> > - If my assumption is correct, the tcsetattr function must not
> >    set the Windows VMIN equivalent to != 0 in the O_NONBLOCK case.
> > 
> > - Also, as mention above, if VTIME > 0, Cygwin's read sets the number of
> >    bytes_to_read to 1.   Weirdly, it does so even if there are more than
> >    1 byte in the inbound queue.  This sounds wrong to me.
> > 
> I think the problem is if the number of bytes requested are more than what

To clarify: number of bytes == VMIN?

> is in the buffer it is going to overlap the read function (because of VTIME)
> and immediately after that CancelIO is called. Contrary to what is mentioned
> in the source code I think CancelIO is actually discarding data...

So far we didn't have that experience.  CancelIO is usually safe
in this regard.

>   /* Use CancelIo rather than PurgeComm (PURGE_RXABORT) since
>          PurgeComm apparently discards in-flight bytes while CancelIo
>          only stops the overlapped IO routine. */
> 
> 
> My suggestion is the following patch:
> 
> diff --git a/winsup/cygwin/fhandler_serial.cc
> b/winsup/cygwin/fhandler_serial.cc
> index 69e5768f6..afa8871bf 100644
> --- a/winsup/cygwin/fhandler_serial.cc
> +++ b/winsup/cygwin/fhandler_serial.cc
> @@ -898,7 +898,11 @@ fhandler_serial::tcsetattr (int action, const struct
> termios *t)
>    {
>      memset (&to, 0, sizeof (to));
> 
> -    if ((vmin_ > 0) && (vtime_ == 0))
> +       if (is_nonblocking())
> +       {
> +               to.ReadIntervalTimeout = MAXDWORD;
> +       }
> +    else if ((vmin_ > 0) && (vtime_ == 0))

What if you switch to !O_NONBLOCK after calling tcsetattr?  The
setting of ReadIntervalTimeout would be lost then.

Either we have to repeat calling SetCommTimeouts every time
we switch mode, or we have to do the above setting temporary
every time we call ReadFile in non blocking mode.


Corinna

-- 
Corinna Vinschen
Cygwin Maintainer
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://cygwin.com/pipermail/cygwin/attachments/20200312/4787ae4e/attachment.sig>