Re: [PATCH RFC 0/11] usbserial: Serial Core port.

[Pete Zaitcev <zaitcev@redhat.com>]

On Fri, 02 Jun 2006 00:03:06 -0300, "Luiz Fernando N.Capitulino" <lcapitulino@mandriva.com.br> wrote:

This looks interesting, although I do not know if it buys us much.
The code seems sane at first view. The private lock inside pl2303
saves you from the most obvious races.

>  The tests I've done so far weren't anything serious: as the mobile supports a
> AT command set, I have used the ones (with minicom) which transfers more data.
> Of course that I also did module load/unload tests, tried to disconnect the
> device while it's transfering data and so on.

Next, it would be nice to test if PPP works, and if getty and shell work
(with getty driving the USB-to-serial adapter).

> +static void serial_send_xchar(struct uart_port *port, char ch)
> +{
> +	USBSERIAL_PORT->serial->type->uart_ops->send_xchar(port, ch);
>  }

I think you just inherited a mistake in usb-serial design. It attempts
to act as an adaptation layer (like, say, USB core itself) instead of
a library like libata. Why can't the UART framework call pl2303?

Also this meaningless obfuscation has to go:

> +#define USBSERIAL_PORT ((struct usb_serial_port *)port)
> +static void pl2303_start_tx(struct uart_port *port)
> +{
> +	struct usb_serial_port *usp = USBSERIAL_PORT;

Greetings,
-- Pete