[PATCH] lib: utils/serial: Support Synopsys DesignWare APB UART

[Bin Meng]

On Fri, May 14, 2021 at 9:59 AM Jessica Clarke <jrtc27 at jrtc27.com> wrote:
>
> On 14 May 2021, at 02:27, Bin Meng <bmeng.cn at gmail.com> wrote:
> >
> > On Fri, May 14, 2021 at 9:19 AM Jessica Clarke <jrtc27 at jrtc27.com> wrote:
> >>
> >> On 14 May 2021, at 02:16, Bin Meng <bmeng.cn at gmail.com> wrote:
> >>>
> >>> Synopsys DesignWare APB UART is seen on the StarFive JH7100 SoC.
> >>> Its programming interface is compatible with the existing 8250
> >>> UART driver. Simply add its compatible string to the driver makes
> >>> it work with the StarFive JH7100 SoC on a BeagleV board.
> >>>
> >>> With this patch, the generic platform firmware can be used out of
> >>> the box on the BeagleV board.
> >>>
> >>> Signed-off-by: Bin Meng <bmeng.cn at gmail.com>
> >>> ---
> >>>
> >>> lib/utils/serial/fdt_serial_uart8250.c | 1 +
> >>> 1 file changed, 1 insertion(+)
> >>>
> >>> diff --git a/lib/utils/serial/fdt_serial_uart8250.c b/lib/utils/serial/fdt_serial_uart8250.c
> >>> index 918193a..36f364c 100644
> >>> --- a/lib/utils/serial/fdt_serial_uart8250.c
> >>> +++ b/lib/utils/serial/fdt_serial_uart8250.c
> >>> @@ -28,6 +28,7 @@ static int serial_uart8250_init(void *fdt, int nodeoff,
> >>> static const struct fdt_match serial_uart8250_match[] = {
> >>>      { .compatible = "ns16550" },
> >>>      { .compatible = "ns16550a" },
> >>> +     { .compatible = "snps,dw-apb-uart" },
> >>
> >> If it’s compatible, why does it not use
> >>
> >>  compatible = “snps,dw-apb-uart”, “ns16550”;
> >>
> >> or similar in its FDT?
> >
> > Good question. By looking at the Linux kernel driver, it is supported
> > as a variant of 8250 (drivers/tty/serial/8250/8250_dw.c).
> >
> > I have not looked into details, but I suspect there are some minor
> > oddities which matter for the kernel. The existing driver is enough to
> > make OpenSBI happy.
>
> It seems there are various additional clocks and resets you may need to handle
> (baudclk is required, apb_pclk is optional and there’s an optional reset too).
> For this particular board the FDT has a them clocks as all fixed-clock, but in
> general this is not a correct implementation of the driver, as the clocks must
> be enabled when attaching.

I don't think OpenSBI needs to handle the clock and reset of each
peripheral in its driver. Such should to be done in the bootloader,
i.e.: U-Boot SPL.

Regards,
Bin