Re: [PATCH v2 3/5] drivers/tty/serial: add driver for the ESP32 UART

From: Max Filippov
Date: Wed Sep 20 2023 - 04:35:10 EST

Next message: Conor Dooley: "Re: [PATCH v2] riscv: dts: thead: set dma-noncoherent to soc bus"
Previous message: kernel test robot: "[tip:core/core] BUILD SUCCESS 1aabbc532413ced293952f8e149ad0a607d6e470"
In reply to: Jiri Slaby: "Re: [PATCH v2 3/5] drivers/tty/serial: add driver for the ESP32 UART"
Next in thread: Max Filippov: "[PATCH v2 5/5] drivers/tty/serial: add ESP32S3 ACM device driver"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Wed, Sep 20, 2023 at 12:22 AM Jiri Slaby <jirislaby@xxxxxxxxxx> wrote:
>
> On 20. 09. 23, 4:26, Max Filippov wrote:
> > Add driver for the UART controllers of the Espressif ESP32 and ESP32S3
> > SoCs. Hardware specification is available at the following URLs:
> ...
> > +static void esp32_uart_rxint(struct uart_port *port)
> > +{
> > + struct tty_port *tty_port = &port->state->port;
> > + u32 rx_fifo_cnt = esp32_uart_rx_fifo_cnt(port);
> > + unsigned long flags;
> > + u32 i;
> > +
> > + if (!rx_fifo_cnt)
> > + return;
> > +
> > + spin_lock_irqsave(&port->lock, flags);
> > +
> > + for (i = 0; i < rx_fifo_cnt; ++i) {
> > + u32 rx = esp32_uart_read(port, UART_FIFO_REG);
> > + u32 brk = 0;
> > +
> > + ++port->icount.rx;
>
> Should yuou increment this only in case of insert_flip_char()?

I don't know. Does port->icount.rx have clearly defined semantics?
I've looked through multiple other serial drivers and there's a lot of
examples of similar pattern.

> > + if (!rx) {
> > + brk = esp32_uart_read(port, UART_INT_ST_REG) &
> > + UART_BRK_DET_INT;
> > + }
> > + if (brk) {
> > + esp32_uart_write(port, UART_INT_CLR_REG, brk);
> > + ++port->icount.brk;
> > + uart_handle_break(port);
> > + } else {
> > + if (uart_handle_sysrq_char(port, (unsigned char)rx))
> > + continue;
> > + tty_insert_flip_char(tty_port, rx, TTY_NORMAL);
> > + }
>
> This is heavy. Is it equivalent to:
> if (!rx && esp32_uart_read(port, UART_INT_ST_REG) &
> UART_BRK_DET_INT) {
> esp32_uart_write(port, UART_INT_CLR_REG, brk);
> ++port->icount.brk;
> uart_handle_break(port);
> continue;
> }
>
> if (uart_handle_sysrq_char(port, (unsigned char)rx))
> continue;
>
> tty_insert_flip_char(tty_port, rx, TTY_NORMAL);
>
> ?

It is equivalent, but I find the version that I used somewhat easier to read.
Maybe this:

if (!rx &&
(esp32_uart_read(port, UART_INT_ST_REG) &
UART_BRK_DET_INT)) {
esp32_uart_write(port, UART_INT_CLR_REG, brk);
++port->icount.brk;
uart_handle_break(port);
} else {
if (uart_handle_sysrq_char(port, (unsigned char)rx))
continue;
tty_insert_flip_char(tty_port, rx, TTY_NORMAL);
}

?

> > + }
> > + spin_unlock_irqrestore(&port->lock, flags);
> > +
> > + tty_flip_buffer_push(tty_port);
> > +}
> ...
> > +static void esp32_uart_put_char_sync(struct uart_port *port, u8 c)
> > +{
> > + unsigned long timeout;
> > +
> > + timeout = jiffies + msecs_to_jiffies(1000);
>
> I.e. plus HZ?

Yes, ok.

> > + while (esp32_uart_tx_fifo_cnt(port) >= ESP32_UART_TX_FIFO_SIZE) {
> > + if (time_after(jiffies, timeout)) {
> > + dev_warn(port->dev, "timeout waiting for TX FIFO\n");
> > + return;
> > + }
> > + cpu_relax();
> > + }
> > + esp32_uart_put_char(port, c);
> > +}
> > +
> > +static void esp32_uart_transmit_buffer(struct uart_port *port)
> > +{
> > + u32 tx_fifo_used = esp32_uart_tx_fifo_cnt(port);
> > +
> > + if (tx_fifo_used < ESP32_UART_TX_FIFO_SIZE) {
>
> Invert the condition and return here?

Ok.

> > + unsigned int pending;
> > + u8 ch;
> > +
> > + pending = uart_port_tx_limited(port, ch,
> > + ESP32_UART_TX_FIFO_SIZE - tx_fifo_used,
> > + true, esp32_uart_put_char(port, ch),
> > + ({}));
> > + if (pending) {
> > + u32 int_ena;
> > +
> > + int_ena = esp32_uart_read(port, UART_INT_ENA_REG);
> > + int_ena |= UART_TXFIFO_EMPTY_INT;
> > + esp32_uart_write(port, UART_INT_ENA_REG, int_ena);
> > + }
> > + }
> > +}
>
>
> > +static irqreturn_t esp32_uart_int(int irq, void *dev_id)
> > +{
> > + struct uart_port *port = dev_id;
> > + u32 status;
> > +
> > + status = esp32_uart_read(port, UART_INT_ST_REG);
> > +
> > + if (status & (UART_RXFIFO_FULL_INT | UART_BRK_DET_INT))
> > + esp32_uart_rxint(port);
> > + if (status & UART_TXFIFO_EMPTY_INT)
> > + esp32_uart_txint(port);
> > +
> > + esp32_uart_write(port, UART_INT_CLR_REG, status);
>
> \n here please.

Ok

> > + return IRQ_RETVAL(status);
> > +}
>
> > +static int esp32_uart_startup(struct uart_port *port)
> > +{
> > + int ret = 0;
> > + unsigned long flags;
> > + struct esp32_port *sport = container_of(port, struct esp32_port, port);
> > +
> > + ret = clk_prepare_enable(sport->clk);
> > + if (ret)
> > + return ret;
> > +
> > + ret = devm_request_irq(port->dev, port->irq, esp32_uart_int, 0,
> > + DRIVER_NAME, port);
> > + if (ret) {
> > + clk_disable_unprepare(sport->clk);
> > + return ret;
> > + }
> > +
> > + spin_lock_irqsave(&port->lock, flags);
> > + esp32_uart_write(port, UART_CONF1_REG,
> > + (1 << UART_RXFIFO_FULL_THRHD_SHIFT) |
>
> BIT() ?
>
> > + (1 << port_variant(port)->txfifo_empty_thrhd_shift));
>
> And here?

These are not logically bits, in both cases I put value 1 into
multiple-bit fields.

> > + esp32_uart_write(port, UART_INT_CLR_REG, UART_RXFIFO_FULL_INT | UART_BRK_DET_INT);
> > + esp32_uart_write(port, UART_INT_ENA_REG, UART_RXFIFO_FULL_INT | UART_BRK_DET_INT);
> > + spin_unlock_irqrestore(&port->lock, flags);
> > +
> > + pr_debug("%s, conf1 = %08x, int_st = %08x, status = %08x\n",
> > + __func__,
> > + esp32_uart_read(port, UART_CONF1_REG),
> > + esp32_uart_read(port, UART_INT_ST_REG),
> > + esp32_uart_read(port, UART_STATUS_REG));
>
> \n here. Is this debug printout somehow useful?

I'll drop it.

> > + return ret;
> > +}
> > +
> > +static void esp32_uart_shutdown(struct uart_port *port)
> > +{
> > + struct esp32_port *sport = container_of(port, struct esp32_port, port);
> > +
> > + esp32_uart_write(port, UART_INT_ENA_REG, 0);
> > + devm_free_irq(port->dev, port->irq, port);
>
> I wonder why to use devm_ after all?

I'll switch to non-devm_ version.

> > + clk_disable_unprepare(sport->clk);
> > +}
> > +
> > +static bool esp32_uart_set_baud(struct uart_port *port, u32 baud)
> > +{
> > + u32 div = port->uartclk / baud;
> > + u32 frag = (port->uartclk * 16) / baud - div * 16;
> > +
> > + if (div <= port_variant(port)->clkdiv_mask) {
> > + esp32_uart_write(port, UART_CLKDIV_REG,
> > + div | FIELD_PREP(UART_CLKDIV_FRAG, frag));
> > + return true;
> > + }
>
> \n

Ok.

> > + return false;
> > +}
> ...
> > +static int __init esp32s3_uart_early_console_setup(struct earlycon_device *device,
> > + const char *options)
> > +{
> > + device->port.private_data = (void *)&esp32s3_variant;
>
> No need to cast, right?

private_data is void *, esp32s3_variant is a constant.

> \n

Ok.

> > + return esp32xx_uart_early_console_setup(device, options);
> > +}
> ...
> > +static int esp32_uart_probe(struct platform_device *pdev)
> > +{
> > + struct device_node *np = pdev->dev.of_node;
> > + static const struct of_device_id *match;
> > + struct uart_port *port;
> > + struct esp32_port *sport;
> > + struct resource *res;
> > + int ret;
> > +
> > + match = of_match_device(esp32_uart_dt_ids, &pdev->dev);
> > + if (!match)
> > + return -ENODEV;
> > +
> > + sport = devm_kzalloc(&pdev->dev, sizeof(*sport), GFP_KERNEL);
> > + if (!sport)
> > + return -ENOMEM;
> > +
> > + port = &sport->port;
> > +
> > + ret = of_alias_get_id(np, "serial");
> > + if (ret < 0) {
> > + dev_err(&pdev->dev, "failed to get alias id, errno %d\n", ret);
> > + return ret;
> > + }
> > + if (ret >= UART_NR) {
> > + dev_err(&pdev->dev, "driver limited to %d serial ports\n", UART_NR);
> > + return -ENOMEM;
> > + }
> > +
> > + port->line = ret;
> > +
> > + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > + if (!res)
> > + return -ENODEV;
> > +
> > + port->mapbase = res->start;
> > + port->membase = devm_ioremap_resource(&pdev->dev, res);
> > + if (IS_ERR(port->membase))
> > + return PTR_ERR(port->membase);
> > +
> > + sport->clk = devm_clk_get(&pdev->dev, NULL);
> > + if (IS_ERR(sport->clk))
> > + return PTR_ERR(sport->clk);
> > +
> > + port->uartclk = clk_get_rate(sport->clk);
> > + port->dev = &pdev->dev;
> > + port->type = PORT_ESP32UART;
> > + port->iotype = UPIO_MEM;
> > + port->irq = platform_get_irq(pdev, 0);
> > + port->ops = &esp32_uart_pops;
> > + port->flags = UPF_BOOT_AUTOCONF;
> > + port->has_sysrq = 1;
> > + port->fifosize = ESP32_UART_TX_FIFO_SIZE;
> > + port->private_data = (void *)match->data;
>
> No need to cast.

This is again a const cast.

> > +
> > + esp32_uart_ports[port->line] = sport;
> > +
> > + platform_set_drvdata(pdev, port);
> > +
> > + return uart_add_one_port(&esp32_uart_reg, port);
> > +}
>
> regards,
> --
> js
> suse labs
>

--
Thanks.
-- Max

Next message: Conor Dooley: "Re: [PATCH v2] riscv: dts: thead: set dma-noncoherent to soc bus"
Previous message: kernel test robot: "[tip:core/core] BUILD SUCCESS 1aabbc532413ced293952f8e149ad0a607d6e470"
In reply to: Jiri Slaby: "Re: [PATCH v2 3/5] drivers/tty/serial: add driver for the ESP32 UART"
Next in thread: Max Filippov: "[PATCH v2 5/5] drivers/tty/serial: add ESP32S3 ACM device driver"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]