Re: [PATCH v3 08/13] tty: serial: qcom-geni-serial: refactor qcom_geni_serial_handle_tx()

[Bartosz Golaszewski <brgl@bgdev.pl>]

On Thu, Nov 24, 2022 at 8:18 AM Jiri Slaby <jirislaby@kernel.org> wrote:
>
> On 23. 11. 22, 12:07, Bartosz Golaszewski wrote:
> > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> >
> > qcom_geni_serial_handle_tx() is pretty big, let's move the code that
> > handles the actual writing of data to a separate function which makes
> > sense in preparation for introducing a dma variant of handle_tx().
> >
> > Let's also shuffle the code a bit, drop unneeded variables and use
> > uart_xmit_advance() instead of handling tail->xmit manually.
> >
> > Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> > ---
> >   drivers/tty/serial/qcom_geni_serial.c | 54 +++++++++++++--------------
> >   1 file changed, 27 insertions(+), 27 deletions(-)
> >
> > diff --git a/drivers/tty/serial/qcom_geni_serial.c b/drivers/tty/serial/qcom_geni_serial.c
> > index 68a1402fbe58..658b6d596f58 100644
> > --- a/drivers/tty/serial/qcom_geni_serial.c
> > +++ b/drivers/tty/serial/qcom_geni_serial.c
> > @@ -704,19 +704,42 @@ static void qcom_geni_serial_start_rx(struct uart_port *uport)
> >       writel(irq_en, uport->membase + SE_GENI_M_IRQ_EN);
> >   }
>
> I know you just shuffle the code, but:
>
> > +static void qcom_geni_serial_send_chunk_fifo(struct uart_port *uport,
> > +                                          unsigned int chunk)
> > +{
> > +     struct qcom_geni_serial_port *port = to_dev_port(uport);
> > +     struct circ_buf *xmit = &uport->state->xmit;
> > +     u8 buf[BYTES_PER_FIFO_WORD];
> > +     size_t remaining = chunk;
>
> Why size_t when the others are uints? Well, BYTES_PER_FIFO_WORD should
> be defined as 4U.

Good point.

>
> > +     unsigned int tx_bytes;
> > +     int c;
> > +
> > +     while (remaining) {
> > +             memset(buf, 0, sizeof(buf));
> > +             tx_bytes = min_t(size_t, remaining, BYTES_PER_FIFO_WORD);
>
> Then, no need for min_t.
>

Same.

> > +
> > +             for (c = 0; c < tx_bytes ; c++) {
> > +                     buf[c] = xmit->buf[xmit->tail];
> > +                     uart_xmit_advance(uport, 1);
> > +             }
> > +
> > +             iowrite32_rep(uport->membase + SE_GENI_TX_FIFOn, buf, 1);
>
> I wonder, why is _rep variant used to transfer a single word? Only to
> hide the cast?
>

Even if - using writel() with a cast doesn't seem to improve the
performance and this one looks prettier IMO.

Bartosz