Received-SPF: pass (google.com: domain of linux-kernel+bounces-101628-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Wed, 13 Mar 2024 10:37:54 -0400
From: Hugo Villeneuve <hugo@hugovil.com>
To: Hugo Villeneuve <hugo@hugovil.com>
Cc: gregkh@linuxfoundation.org, jirislaby@kernel.org,
 linux-kernel@vger.kernel.org, linux-serial@vger.kernel.org, Hugo Villeneuve
 <hvilleneuve@dimonoff.com>
Message-Id: <20240313103754.6c33d160f5d69427a29a3e2b@hugovil.com>
In-Reply-To: <20240307161828.118495-4-hugo@hugovil.com>
References: <20240307161828.118495-1-hugo@hugovil.com>
	<20240307161828.118495-4-hugo@hugovil.com>
Precedence: bulk
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Subject: Re: [PATCH v2 3/4] serial: sc16is7xx: split into core and I2C/SPI
 parts (sc16is7xx_lines)

On Thu,  7 Mar 2024 11:18:27 -0500
Hugo Villeneuve <hugo@hugovil.com> wrote:

> From: Hugo Villeneuve <hvilleneuve@dimonoff.com>
> 
> Before, sc16is7xx_lines was checked for a free (zero) bit first, and then
> later it was set only if UART port registration succeeded.
> 
> Now that sc16is7xx_lines is shared for the I2C and SPI drivers, make sure
> it is reserved and modified atomically, and use a new variable to hold the
> status of UART port regisration.
> 
> Remove need to check for previous port registration in sc16is7xx_remove(),
> because if sc16is7xx_probe() succeeded, we are guaranteed to have
> successfully registered both ports.
> 
> Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
> ---
>  drivers/tty/serial/sc16is7xx.c | 22 +++++++++++++++++-----
>  1 file changed, 17 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/tty/serial/sc16is7xx.c b/drivers/tty/serial/sc16is7xx.c
> index 5b53c88b7133..d81aad1b201d 100644
> --- a/drivers/tty/serial/sc16is7xx.c
> +++ b/drivers/tty/serial/sc16is7xx.c
> @@ -343,6 +343,8 @@ struct sc16is7xx_port {
>  
>  static DECLARE_BITMAP(sc16is7xx_lines, SC16IS7XX_MAX_DEVS);
>  
> +static DEFINE_MUTEX(sc16is7xx_lines_lock); /* For probe atomic line reservation. */
> +
>  static struct uart_driver sc16is7xx_uart = {
>  	.owner		= THIS_MODULE,
>  	.driver_name    = SC16IS7XX_NAME,
> @@ -1468,6 +1470,7 @@ int sc16is7xx_probe(struct device *dev, const struct sc16is7xx_devtype *devtype,
>  	u32 uartclk = 0;
>  	int i, ret;
>  	struct sc16is7xx_port *s;
> +	bool port_registered[SC16IS7XX_MAX_PORTS];
>  
>  	for (i = 0; i < devtype->nr_uart; i++)
>  		if (IS_ERR(regmaps[i]))
> @@ -1532,14 +1535,21 @@ int sc16is7xx_probe(struct device *dev, const struct sc16is7xx_devtype *devtype,
>  	regmap_write(regmaps[0], SC16IS7XX_IOCONTROL_REG,
>  		     SC16IS7XX_IOCONTROL_SRESET_BIT);
>  
> +	memset(port_registered, 0, sizeof(port_registered));
> +
>  	for (i = 0; i < devtype->nr_uart; ++i) {
> +		mutex_lock(&sc16is7xx_lines_lock);
>  		s->p[i].port.line = find_first_zero_bit(sc16is7xx_lines,
>  							SC16IS7XX_MAX_DEVS);
>  		if (s->p[i].port.line >= SC16IS7XX_MAX_DEVS) {
>  			ret = -ERANGE;
> +			mutex_unlock(&sc16is7xx_lines_lock);
>  			goto out_ports;
>  		}
>  
> +		set_bit(s->p[i].port.line, sc16is7xx_lines);
> +		mutex_unlock(&sc16is7xx_lines_lock);
> +
>  		/* Initialize port data */
>  		s->p[i].port.dev	= dev;
>  		s->p[i].port.irq	= irq;
> @@ -1584,7 +1594,7 @@ int sc16is7xx_probe(struct device *dev, const struct sc16is7xx_devtype *devtype,
>  		if (ret)
>  			goto out_ports;
>  
> -		set_bit(s->p[i].port.line, sc16is7xx_lines);
> +		port_registered[i] = true;
>  
>  		/* Enable EFR */
>  		sc16is7xx_port_write(&s->p[i].port, SC16IS7XX_LCR_REG,
> @@ -1642,9 +1652,11 @@ int sc16is7xx_probe(struct device *dev, const struct sc16is7xx_devtype *devtype,
>  #endif
>  
>  out_ports:
> -	for (i = 0; i < devtype->nr_uart; i++)
> -		if (test_and_clear_bit(s->p[i].port.line, sc16is7xx_lines))
> +	for (i = 0; i < devtype->nr_uart; i++) {
> +		clear_bit(s->p[i].port.line, sc16is7xx_lines);
> +		if (port_registered[i])

Hi Andy,
I just realised that since we no longer need to search if a
bit is set, it is possible to simplify sc16is7xx_lines allocation
by using the IDA framework, as you suggested a few months ago.

I will send a V3 with this change.

Hugo Villeneuve


>  			uart_remove_one_port(&sc16is7xx_uart, &s->p[i].port);
> +	}
>  
>  	kthread_stop(s->kworker_task);
>  
> @@ -1667,8 +1679,8 @@ void sc16is7xx_remove(struct device *dev)
>  
>  	for (i = 0; i < s->devtype->nr_uart; i++) {
>  		kthread_cancel_delayed_work_sync(&s->p[i].ms_work);
> -		if (test_and_clear_bit(s->p[i].port.line, sc16is7xx_lines))
> -			uart_remove_one_port(&sc16is7xx_uart, &s->p[i].port);
> +		clear_bit(s->p[i].port.line, sc16is7xx_lines);
> +		uart_remove_one_port(&sc16is7xx_uart, &s->p[i].port);
>  		sc16is7xx_power(&s->p[i].port, 0);
>  	}
>  
> -- 
> 2.39.2
> 
>