Re: [Ftdi-usb-sio-devel] Question about FlowControl.

[Bill R. <bil...@gm...>]

On Sat, Jan 31, 2009 at 4:15 AM, Toan Pham <tph...@gm...> wrote:

> Bill,
>
> I have a question about flow control in the ftdi driver.
> I notice that when I use serial port without flow control, the
> ftdi driver still does some kind of flow control with the ftdi chip.
> This is found in the set_mctrl, clear_mctrl, and update_mctrl functions.


i don't know about those functions. Ian Abbott implemented them.

http://www.mail-archive.com/bk-...@vg.../msg07158.html

Looks like using them can make a massive difference to performance.

The best place to look will probably be the usb_serial stuff or further up
the stack in the tty driver.



>
>
> Is there a documentation on how flow control is implemented in this driver?
>
> Also, in the function update_mctrl, when something goes bad
> you disconnect the usb device, as seen in the code below.
> I've been experiencing alot of disconnections at this point,
> maybe due to noise in the bus; but I've tried shorter cable
> length, and disconnection still happen.  Would you please
> explain why we disconnect here, and how often we do
> flow control request with the chip?  thank you.



I assume that if an error is seen the assumption is made that the usb
connection needs to be dropped.

You could try removing the usb_disconnect but to be careful I suggest you
dig around and see which error codes returned by usb_control_msg you should
and shouldn't drop the connection on.

Of course be prepared for panics or something if you try this!



>
>
>
>
>
> REFERENCE:
>
> ---------------------------------------------------------------------------------
> static int update_mctrl(struct usb_serial_port *port, unsigned int
> set, unsigned int clear)
> {
>        struct ftdi_private *priv = usb_get_serial_port_data(port);
>        char *buf;
>        unsigned urb_value;
>        int rv;
>
>        if (((set | clear) & (TIOCM_DTR | TIOCM_RTS)) == 0) {
>                dbg("%s - DTR|RTS not being set|cleared", __FUNCTION__);
>                return 0;       /* no change */
>        }
>
>        buf = kmalloc(1, GFP_NOIO);
>        if (!buf) {
>                return -ENOMEM;
>        }
>
>        clear &= ~set;  /* 'set' takes precedence over 'clear' */
>        urb_value = 0;
>        if (clear & TIOCM_DTR)
>                urb_value |= FTDI_SIO_SET_DTR_LOW;
>        if (clear & TIOCM_RTS)
>                urb_value |= FTDI_SIO_SET_RTS_LOW;
>        if (set & TIOCM_DTR)
>                urb_value |= FTDI_SIO_SET_DTR_HIGH;
>        if (set & TIOCM_RTS)
>                urb_value |= FTDI_SIO_SET_RTS_HIGH;
>        rv = usb_control_msg(port->serial->dev,
>                               usb_sndctrlpipe(port->serial->dev, 0),
>                               FTDI_SIO_SET_MODEM_CTRL_REQUEST,
>                               FTDI_SIO_SET_MODEM_CTRL_REQUEST_TYPE,
>                               urb_value, priv->interface,
>                               buf, 0, WDR_TIMEOUT);
>
>        kfree(buf);
>        if (rv < 0) {
>                err("%s Error from MODEM_CTRL urb: DTR %s, RTS %s",
>                                __FUNCTION__,
>                                (set & TIOCM_DTR) ? "HIGH" :
>                                (clear & TIOCM_DTR) ? "LOW" : "unchanged",
>                                (set & TIOCM_RTS) ? "HIGH" :
>                                (clear & TIOCM_RTS) ? "LOW" : "unchanged");
>                usb_disconnect(port);  /*  THIS IS WHERE I OFFEN GET
> DISCONNECTIONS,
> i AM NOT SURE WHAT WENT BAD, HOST CONTROLLER, HUB, DEVICE?*/
>        } else {
>                dbg("%s - DTR %s, RTS %s", __FUNCTION__,
>                                (set & TIOCM_DTR) ? "HIGH" :
>                                (clear & TIOCM_DTR) ? "LOW" : "unchanged",
>                                (set & TIOCM_RTS) ? "HIGH" :
>                                (clear & TIOCM_RTS) ? "LOW" : "unchanged");
>                priv->last_dtr_rts = (priv->last_dtr_rts & ~clear) | set;
>        }
>        return rv;
> }
>
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by:
> SourcForge Community
> SourceForge wants to tell your story.
> http://p.sf.net/sfu/sf-spreadtheword
> _______________________________________________
> Ftdi-usb-sio-devel mailing list
> Ftd...@li...
> https://lists.sourceforge.net/lists/listinfo/ftdi-usb-sio-devel
>