Re: [PATCH 4/5] Add driver for SUNIX Multi-I/O board

From: Enrico Weigelt, metux IT consult
Date: Wed Mar 06 2019 - 22:30:13 EST

Next message: Eric Wong: "Re: [RFC] LKML Archive in Maildir Format"
Previous message: Qii Wang: "Re: [PATCH v4 3/3] i2c: mediatek: Add i2c support for MediaTek MT8183"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On 27.02.19 08:18, Morris Ku wrote:

Hi,

first of all: the code is pretty unreadable. I doubt that anybody here
seriously likes to review that (nor take it into mainline).

Formatting should be consistent - see other drivers.

./scripts/checkpatch.pl is your friend. You really shut run it and fix
everything it complaints before posting patches. (some maintainers can
get angry if you don't ;-)

> Support SUNIX serial board.
>
> ---
> char/snx/snx_serial.c | 4771 +++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 4771 insertions(+)
> create mode 100644 char/snx/snx_serial.c
>
> diff --git a/char/snx/snx_serial.c b/char/snx/snx_serial.c
> new file mode 100644
> index 00000000..94caac1a
> --- /dev/null
> +++ b/char/snx/snx_serial.c
> @@ -0,0 +1,4771 @@
> +#include "snx_common.h"
> +#include "driver_extd.h"
> +
> +#define SNX_ioctl_DBG 0
> +#define EEPROM_ACCESS_DELAY_COUNT 100000

is this eeprom stuff really specific to the serial ports ?
if not, it probably should go to a separate file, which is called by all
the others, IMHO.

> +
> +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37))
> + static DEFINE_SEMAPHORE(ser_port_sem);
> +#else
> + static DECLARE_MUTEX(ser_port_sem);
> +#endif

Drop that. We have only one kernel version here, the current one.

> +
> +
> +#define SNX_HIGH_BITS_OFFSET ((sizeof(long)-sizeof(int))*8)
> +#define sunix_ser_users(state) ((state)->count + ((state)->info ? (state)->info->blocked_open : 0))
> +
> +
> +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3, 7, 0))
> +static struct tty_port snx_service_port;
> +#endif

are you really sure this has to be a global field, instead of allocated
per-device ?

> +
> +
> +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3, 7, 0))
> +
> +struct serial_uart_config {
> + char *name;
> + int dfl_xmit_fifo_size;
> + int flags;
> +};
> +#endif
> +
> +static const struct serial_uart_config snx_uart_config[PORT_SER_MAX_UART + 1] = {

why +1 ?

maybe PORT_SER_MAX_UART and use ARRAY_SIZE() macro instead.

> + { "unknown", 1, 0 },
> + { "8250", 1, 0 },
> + { "16450", 1, 0 },
> + { "16550", 1, 0 },
> + { "16550A", 16, UART_CLEAR_FIFO | UART_USE_FIFO },
> + { "Cirrus", 1, 0 },
> + { "ST16650", 1, 0 },
> + { "ST16650V2", 32, UART_CLEAR_FIFO | UART_USE_FIFO },
> + { "TI16750", 64, UART_CLEAR_FIFO | UART_USE_FIFO },
> +};
>
> +
> +
> +#if (LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 0))
> +static int sunix_ser_refcount;
> +static struct tty_struct *sunix_ser_tty[SNX_SER_TOTAL_MAX + 1];
> +static struct termios *sunix_ser_termios[SNX_SER_TOTAL_MAX + 1];
> +static struct termios *sunix_ser_termios_locked[SNX_SER_TOTAL_MAX + 1];
> +#endif

obsolete. drop that.

> +
> +
> +static _INLINE_ void snx_ser_handle_cts_change(struct snx_ser_port *, unsigned int);

what exactly does _INLINE_ suppose to mean ?

<snip>

> +extern int sunix_ser_interrupt(struct sunix_board *, struct sunix_ser_port *);

why "extern" here ? where is that function coming from ?

<snip>

> +static unsigned char READ_INTERRUPT_VECTOR_BYTE(struct sunix_ser_port *);

I doubt that upper case function names fit in the kernel coding style.

<snip>

> + case SNX_SER_DUMP_PORT_INFO:
> + {
> + int i;
> + struct snx_ser_port_info snx_port_info[SNX_SER_TOTAL_MAX];
> + struct snx_ser_port *sdn = NULL;
> +
> + memset(snx_port_info, 0, (sizeof(struct snx_ser_port_info) * SNX_SER_TOTAL_MAX));
> +
> + if (line == 0) {
> + for (i = 0; i < SNX_SER_TOTAL_MAX; i++) {
> + sdn = (struct snx_ser_port *) &sunix_ser_table[i];
> +
> + memcpy(&snx_port_info[i].board_name_info[0], &sdn->pb_info.board_name[0], SNX_BOARDNAME_LENGTH);
> +
> + snx_port_info[i].bus_number_info = sdn->bus_number;
> + snx_port_info[i].dev_number_info = sdn->dev_number;
> + snx_port_info[i].port_info = sdn->line;
> + snx_port_info[i].base_info = sdn->iobase;
> + snx_port_info[i].irq_info = sdn->irq;
> + }
> +
> + if (copy_to_user((void *)arg, snx_port_info, SNX_SER_TOTAL_MAX * sizeof(struct snx_ser_port_info))) {
> + ret = -EFAULT;
> + } else {
> + ret = 0;
> + }
> + } else {
> + ret = 0;
> + }
> +
> + ret = 0;
> + break;
> + }
> +
> + case SNX_SER_DUMP_DRIVER_VER:
> + {
> + char driver_ver[SNX_DRIVERVERSION_LENGTH];
> + memset(driver_ver, 0, (sizeof(char) * SNX_DRIVERVERSION_LENGTH));
> +
> + if (line == 0) {
> +
> + memcpy(&driver_ver[0], SNX_DRIVER_VERSION, sizeof(SNX_DRIVER_VERSION));
> +
> + if (copy_to_user((void *)arg, &driver_ver, (sizeof(char) * SNX_DRIVERVERSION_LENGTH))) {
> + ret = -EFAULT;
> + } else {
> + ret = 0;
> + }
> + } else {
> + ret = 0;
> + }
> +
> + break;
> + }
> +
> +
> + case SNX_COMM_GET_BOARD_CNT:
> + {
> + SNX_DRVR_BOARD_CNT gb;
> +
> + memset(&gb, 0, (sizeof(SNX_DRVR_BOARD_CNT)));
> +
> + gb.cnt = snx_board_count;
> +
> + if (copy_to_user((void *)arg, &gb, (sizeof(SNX_DRVR_BOARD_CNT)))) {
> + ret = -EFAULT;
> + } else {
> + ret = 0;
> + }
> +
> + break;
> + }
> +
> + case SNX_COMM_GET_BOARD_INFO:
> + {
> + struct sunix_board *sb = NULL;
> + SNX_DRVR_BOARD_INFO board_info;
> +
> + memset(&board_info, 0, (sizeof(SNX_DRVR_BOARD_INFO)));
> +
> + if (copy_from_user(&board_info, (void *)arg, (sizeof(SNX_DRVR_BOARD_INFO)))) {
> + ret = -EFAULT;
> + } else {
> + ret = 0;
> + }
> +
> + sb = &sunix_board_table[board_info.board_id - 1];
> +
> + board_info.subvender_id = sb->pb_info.sub_vendor_id;
> + board_info.subsystem_id = sb->pb_info.sub_device_id;
> + board_info.oem_id = sb->oem_id;
> + board_info.uart_cnt = sb->uart_cnt;
> + board_info.gpio_chl_cnt = sb->gpio_chl_cnt;
> + board_info.board_uart_type = sb->board_uart_type;
> + board_info.board_gpio_type = sb->board_gpio_type;
> +
> + if (copy_to_user((void *)arg, &board_info, (sizeof(SNX_DRVR_BOARD_INFO)))) {
> + ret = -EFAULT;
> + } else {
> + ret = 0;
> + }
> + break;
> + }
> +

what exactly is that for ?

> + case SNX_GPIO_GET:

gpio stuff doesn't belong into a serial / tty.
for that we have the gpio subsystem. (see drivers/gpio/*)

> + case SNX_UART_GET_TYPE:
> + {
> + struct sunix_board *sb = NULL;
> + SNX_DRVR_UART_GET_TYPE gb;
> +
> + int bar3_base_Address;
> + int bar1_base_Address;
> +
> + int bar3_byte5;
> + int uart_type;
> + int RS422state;
> + int AHDCstate;
> +
> + memset(&gb, 0, (sizeof(SNX_DRVR_UART_GET_TYPE)));
> +
> + if (copy_from_user(&gb, (void *)arg, (sizeof(SNX_DRVR_UART_GET_TYPE)))) {
> + ret = -EFAULT;
> + } else {
> + ret = 0;
> + }
> +
> + sb = &sunix_board_table[gb.board_id - 1];
> +
> + bar3_base_Address = pci_resource_start(sb->pdev, 3);
> + bar1_base_Address = pci_resource_start(sb->pdev, 1);
> +
> + bar3_byte5 = inb(bar3_base_Address + 5);
> + uart_type = (bar3_byte5 & 0xC0) >> 6;
> +
> + if (gb.uart_num <= 4) {
> + AHDCstate = inb(bar3_base_Address + 2) & 0x0F & (0x01 << ((gb.uart_num - 1) % 4));
> + RS422state = inb(bar3_base_Address + 3) & 0xF0 & (0x10 << ((gb.uart_num - 1) % 4));
> + } else if (gb.uart_num <= 8) {
> + AHDCstate = inb(bar1_base_Address + 0x32) & 0x0F & (0x01 << ((gb.uart_num - 1) % 4));
> + RS422state = inb(bar1_base_Address + 0x33) & 0xF0 & (0x10 << ((gb.uart_num - 1) % 4));
> + } else if (gb.uart_num <= 12) {
> + AHDCstate = inb(bar1_base_Address + 0x32 + 0x40) & 0x0F & (0x01 << ((gb.uart_num - 1) % 4));
> + RS422state = inb(bar1_base_Address + 0x33 + 0x40) & 0xF0 & (0x10 << ((gb.uart_num - 1) % 4));
> + } else if (gb.uart_num <= 16) {
> + AHDCstate = inb(bar1_base_Address + 0x32 + 0x80) & 0x0F & (0x01 << ((gb.uart_num - 1) % 4));
> + RS422state = inb(bar1_base_Address + 0x33 + 0x80) & 0xF0 & (0x10 << ((gb.uart_num - 1) % 4));
> + } else {
> + //cmn_err(CE_NOTE, "WARNING : we get an incorrect port number (port = %d)!",gb.uart_num);
> + break;
> + }
> +
> + switch (uart_type) {
> + case 0: // RS-232
> + {
> + gb.uart_type = 0;
> + break;
> + }
> + case 1: // RS-422/485
> + {
> + if (AHDCstate && RS422state) {
> + gb.uart_type = 3;
> + } else if (AHDCstate && !RS422state) {
> + gb.uart_type = 2;
> + } else if (!AHDCstate && RS422state) {
> + gb.uart_type = 1;
> + } else {
> + gb.uart_type = -1;
> + }
> + break;
> + }
> + case 2:
> + {
> + if (AHDCstate && RS422state) {
> + gb.uart_type = 3;
> + } else if (AHDCstate && !RS422state) {
> + gb.uart_type = 2;
> + } else if (!AHDCstate && RS422state) {
> + gb.uart_type = 1;
> + } else if (!AHDCstate && !RS422state) {
> + gb.uart_type = 0;
> + } else {
> + gb.uart_type = -1;
> + }
> + break;
> + }
> + }
> +
> + if (copy_to_user((void *)arg, &gb, (sizeof(SNX_DRVR_UART_GET_TYPE)))) {
> + ret = -EFAULT;
> + } else {
> + ret = 0;
> + }
> +

what is that for ?

> + case SNX_UART_GET_ACS:
> + {

what is an "ACS" ?

In general: don't arbitrarily introduce new ioctl()s, especially not
device specific ones - unless there is a damn good reason.

--mtx
--
Enrico Weigelt, metux IT consult
Free software and Linux embedded engineering
info@xxxxxxxxx -- +49-151-27565287

Next message: Eric Wong: "Re: [RFC] LKML Archive in Maildir Format"
Previous message: Qii Wang: "Re: [PATCH v4 3/3] i2c: mediatek: Add i2c support for MediaTek MT8183"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]