Re: [PATCH v5 2/2] serial:sunplus-uart:Add Sunplus SoC UART Driver

From: hammer hsieh
Date: Fri Dec 24 2021 - 02:16:56 EST


Hi, Greg KH :

In patch v1 coding quite mess, it is almost 2000 LOCs.
For down size code under 1000 LOCs, I decide to drop DMA function code
after patch v3.
I think that's the biggest difference compared with 8250.
Without DMA function, like you said it looks like 8250 variant.
I think I should put DMA function back in next submit.

Another question for why I need PORT_SLUNPLUS ?
I just check many other uart driver, almost all driver define their
own PORT number.
Actually, I didn't know about it.
Maybe some device like bluetooth(use uart port) need autoconfig.
Then it will call ioctl with TIOCSERCONFIG.
I don't have tool for calling type/config/request/release/verify.

Regards,
Hammer Hsieh

Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> 於 2021年12月21日 週二 下午4:21寫道:
>
> On Tue, Dec 21, 2021 at 04:14:16PM +0800, hammer hsieh wrote:
> > Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> 於 2021年12月20日 週一 下午11:51寫道:
> > >
> > > On Mon, Dec 13, 2021 at 03:10:07PM +0800, Hammer Hsieh wrote:
> > > > +/* Register offsets */
> > > > +#define SUP_UART_DATA 0x00
> > > > +#define SUP_UART_LSR 0x04
> > > > +#define SUP_UART_MSR 0x08
> > > > +#define SUP_UART_LCR 0x0C
> > > > +#define SUP_UART_MCR 0x10
> > > > +#define SUP_UART_DIV_L 0x14
> > > > +#define SUP_UART_DIV_H 0x18
> > > > +#define SUP_UART_ISC 0x1C
> > > > +#define SUP_UART_TX_RESIDUE 0x20
> > > > +#define SUP_UART_RX_RESIDUE 0x24
> > > > +
> > > > +/* Line Status Register bits */
> > > > +#define SUP_UART_LSR_TXE BIT(6) /* tx empty */
> > > > +#define SUP_UART_LSR_BC BIT(5) /* break condition status */
> > > > +#define SUP_UART_LSR_FE BIT(4) /* frame error status */
> > > > +#define SUP_UART_LSR_OE BIT(3) /* overrun error status */
> > > > +#define SUP_UART_LSR_PE BIT(2) /* parity error status */
> > > > +#define SUP_UART_LSR_RX BIT(1) /* 1: receive fifo not empty */
> > > > +#define SUP_UART_LSR_TX BIT(0) /* 1: transmit fifo is not full */
> > > > +#define SUP_UART_LSR_TX_NOT_FULL 1
> > > > +#define SUP_UART_LSR_BRK_ERROR_BITS GENMASK(5, 2)
> > > > +
> > > > +/* Line Control Register bits */
> > > > +#define SUP_UART_LCR_BC BIT(5) /* break condition select */
> > > > +#define SUP_UART_LCR_PR BIT(4) /* parity bit polarity select */
> > > > +#define SUP_UART_LCR_PE BIT(3) /* parity bit enable */
> > > > +#define SUP_UART_LCR_ST BIT(2) /* stop bits select */
> > > > +#define SUP_UART_LCR_WL5 0x00 /* word length 5 */
> > > > +#define SUP_UART_LCR_WL6 0x01 /* word length 6 */
> > > > +#define SUP_UART_LCR_WL7 0x02 /* word length 7 */
> > > > +#define SUP_UART_LCR_WL8 0x03 /* word length 8 (default) */
> > > > +
> > > > +/* Modem Control Register bits */
> > > > +#define SUP_UART_MCR_LB BIT(4) /* Loopback mode */
> > > > +#define SUP_UART_MCR_RI BIT(3) /* ring indicator */
> > > > +#define SUP_UART_MCR_DCD BIT(2) /* data carrier detect */
> > > > +#define SUP_UART_MCR_RTS BIT(1) /* request to send */
> > > > +#define SUP_UART_MCR_DTS BIT(0) /* data terminal ready */
> > > > +
> > > > +/* Interrupt Status/Control Register bits */
> > > > +#define SUP_UART_ISC_RXM BIT(5) /* RX interrupt enable */
> > > > +#define SUP_UART_ISC_TXM BIT(4) /* TX interrupt enable */
> > > > +#define SUP_UART_ISC_RX BIT(1) /* RX interrupt status */
> > > > +#define SUP_UART_ISC_TX BIT(0) /* TX interrupt status */
> > > > +
> > > > +#define SUP_DUMMY_READ BIT(16) /* drop bytes received on a !CREAD port */
> > > > +#define SUP_UART_NR 5
> > >
> > > Aren't most of these defines already in the kernel header files? Why
> > > create them again?
> > >
> >
> > If for reduce code.
> > I can add #include<linux/serial_reg.h>
> > And remove some overlap define name.
> >
> > #define SUP_UART_LCR_PR -> UART_LCR_EPAR
> > #define SUP_UART_LCR_PE -> UART_LCR_PARITY
> > #define SUP_UART_LCR_ST -> UART_LCR_STOP
> > #define SUP_UART_LCR_WL5 -> UART_LCR_WLEN5
> > #define SUP_UART_LCR_WL6 -> UART_LCR_WLEN6
> > #define SUP_UART_LCR_WL7 -> UART_LCR_WLEN7
> > #define SUP_UART_LCR_WL8 -> UART_LCR_WLEN8
> >
> > #define SUP_UART_MCR_LB -> UART_MCR_LOOP
> > #define SUP_UART_MCR_RI -> UART_MCR_OUT2 ?
> > #define SUP_UART_MCR_DCD -> UART_MCR_OUT1 ?
> > #define SUP_UART_MCR_RTS -> UART_MCR_RTS
> > #define SUP_UART_MCR_DTS -> UART_MCR_DTR
> >
> > But the rest define didn't match internal #include<linux/serial_reg.h>
> > , those define still need to keep.
> > Some use SUP_xxxx specific define.
> > Some use internal #include<linux/serial_reg.h>, it is strange.
>
> Do not duplicate defines that we already have for the same hardware
> type.
>
> And again, why is this not a normal serial driver for the existing UART
> types as this hardware is obviously an 8250 variant?
>
> thanks,
>
> greg k-h