Re: [PATCH 18/25] serial: sh-sci: Prepare for multiple clocks and baud rate generators

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

On Friday 20 November 2015 16:17:13 Geert Uytterhoeven wrote:
> On Fri, Nov 20, 2015 at 3:47 PM, Laurent Pinchart wrote:
> > On Friday 20 November 2015 08:52:19 Geert Uytterhoeven wrote:
> >> On Thu, Nov 19, 2015 at 10:04 PM, Laurent Pinchart wrote:
> >> > On Thursday 19 November 2015 19:38:57 Geert Uytterhoeven wrote:
> >> >> Refactor the clock and baud rate parameter code to ease adding support
> >> >> for multiple clocks and baud rate generators later.
> >> >> sci_scbrr_calc() now returns the bit rate error, so it can be compared
> >> >> to the bit rate error for other baud rate generators.
> >> >> 
> >> >> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> >> >> ---
> >> >> 
> >> >>  drivers/tty/serial/sh-sci.c | 176
> >> >>  +++++++++++++++++++++++++-------------
> >> >>  1 file changed, 120 insertions(+), 56 deletions(-)
> >> >> 
> >> >> diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
> >> >> index 726c96d5a511c222..12800e52f41953dc 100644
> >> >> --- a/drivers/tty/serial/sh-sci.c
> >> >> +++ b/drivers/tty/serial/sh-sci.c
> >> >> @@ -2252,33 +2301,48 @@ static struct uart_ops sci_uart_ops = {
> >> >> 
> >> >>  static int sci_init_clocks(struct sci_port *sci_port, struct device
> >> >>  *dev)
> >> >>  {
> >> >> 
> >> >> -     /* Get the SCI functional clock. It's called "fck" on ARM. */
> >> >> -     sci_port->fclk = devm_clk_get(dev, "fck");
> >> >> -     if (PTR_ERR(sci_port->fclk) = -EPROBE_DEFER)
> >> >> -             return -EPROBE_DEFER;
> >> >> -     if (!IS_ERR(sci_port->fclk))
> >> >> -             return 0;
> >> >> +     const char *clk_names[] = {
> >> >> +             [SCI_FCK] = "fck",
> >> >> +     };
> >> >> +     struct clk *clk;
> >> >> +     unsigned int i;
> >> >> 
> >> >> -     /*
> >> >> -      * But it used to be called "sci_ick", and we need to maintain
> >> >> DT
> >> >> -      * backward compatibility.
> >> >> -      */
> >> >> -     sci_port->fclk = devm_clk_get(dev, "sci_ick");
> >> >> -     if (PTR_ERR(sci_port->fclk) = -EPROBE_DEFER)
> >> >> -             return -EPROBE_DEFER;
> >> >> -     if (!IS_ERR(sci_port->fclk))
> >> >> -             return 0;
> >> >> +     for (i = 0; i < SCI_NUM_CLKS; i++) {
> >> >> +             clk = devm_clk_get(dev, clk_names[i]);
> >> >> +             if (PTR_ERR(clk) = -EPROBE_DEFER)
> >> >> +                     return -EPROBE_DEFER;
> >> >> 
> >> >> -     /*
> >> >> -      * Not all SH platforms declare a clock lookup entry for SCI
> >> >> devices,
> >> >> -      * in which case we need to get the global "peripheral_clk"
> >> >> clock.
> >> >> -      */
> >> >> -     sci_port->fclk = devm_clk_get(dev, "peripheral_clk");
> >> >> -     if (!IS_ERR(sci_port->fclk))
> >> >> -             return 0;
> >> >> +             if (IS_ERR(clk) && i = SCI_FCK) {
> >> >> +                     /*
> >> >> +                      * "fck" used to be called "sci_ick", and we
> >> >> need
> >> >> to
> >> >> +                      * maintain DT backward compatibility.
> >> >> +                      */
> >> >> +                     clk = devm_clk_get(dev, "sci_ick");
> >> >> +                     if (PTR_ERR(clk) = -EPROBE_DEFER)
> >> >> +                             return -EPROBE_DEFER;
> >> >> +
> >> >> +                     if (!IS_ERR(clk))
> >> >> +                             goto found;
> >> >> +
> >> >> +                     /*
> >> >> +                      * Not all SH platforms declare a clock lookup
> >> >> entry
> >> >> +                      * for SCI devices, in which case we need to get
> >> >> the
> >> >> +                      * global "peripheral_clk" clock.
> >> >> +                      */
> >> >> +                     clk = devm_clk_get(dev, "peripheral_clk");
> >> >> +                     if (!IS_ERR(clk))
> >> >> +                             goto found;
> >> >> +
> >> >> +                     dev_err(dev, "failed to get functional
> >> >> clock\n");
> >> >> +                     return PTR_ERR(clk);
> >> >> +             }
> >> >> 
> >> >> -     dev_err(dev, "failed to get functional clock\n");
> >> >> -     return PTR_ERR(sci_port->fclk);
> >> >> +found:
> >> >> +             if (!IS_ERR(clk))
> >> >> +                     dev_dbg(dev, "clk %u is %pC rate %pCr\n", i,
> >> >> clk,
> >> >> clk);
> >> >> +             sci_port->clks[i] = IS_ERR(clk) ? NULL : clk;
> >> > 
> >> > Isn't it an issue that we can't tell apart the case where there is no
> >> > clock specified in DT and the case where we can't get the clock due to
> >> > another error ?
> >> 
> >> All failures here are for optional clocks.
> >> If the real failure is that the clock wasn't specified (or misspelled) in
> >> DT, it should have been detected during the integration phase.
> > 
> > There could be cases where the clock is correctly specified in DT but
> > can't be retrieved due to a runtime error. I suppose that's mostly
> > theoretical in our case though. Maybe a dev_dbg for the error case could
> > be useful too ? Can we tell the case where the clock is not specified in
> > DT apart from other errors (-EPROBE_DEFER aside as that case is already
> > handled) ?
> 
> If the clock is not in clock/clock-names in DT, the error is definitely
> -ENOENT.
> 
> If the clock is specified in DT, it has a phandle to a clock node. If that
> clock hasn't been instantiated yet, the error is EPROBE_DEFER.
> Which means there are no other possible error values, right?

There could be other errors returned from __of_clk_get_from_provider if the 
clock provider get method returns an error or if __clk_create_clk() fails. The 
latter only returns -ENOMEM so we'd have worse issues anyway. I agree that a 
clock provider get failure shouldn't be a common case.

-- 
Regards,

Laurent Pinchart