Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
Message-ID: <cf593ff3-bf57-ccd3-9a25-b28cc604d6f4@baylibre.com>
Date:   Mon, 9 May 2022 09:36:40 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.7.0
Subject: Re: [PATCH v1] serial: meson: acquire port->lock in startup()
Content-Language: en-US
To:     John Ogness <john.ogness@linutronix.de>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Kevin Hilman <khilman@baylibre.com>
Cc:     Petr Mladek <pmladek@suse.com>,
        Sergey Senozhatsky <senozhatsky@chromium.org>,
        Steven Rostedt <rostedt@goodmis.org>,
        linux-kernel@vger.kernel.org, Jiri Slaby <jirislaby@kernel.org>,
        Jerome Brunet <jbrunet@baylibre.com>,
        Martin Blumenstingl <martin.blumenstingl@googlemail.com>,
        linux-serial@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
        linux-amlogic@lists.infradead.org,
        Marek Szyprowski <m.szyprowski@samsung.com>
References: <20220508103547.626355-1-john.ogness@linutronix.de>
From:   Neil Armstrong <narmstrong@baylibre.com>
Organization: Baylibre
In-Reply-To: <20220508103547.626355-1-john.ogness@linutronix.de>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 08/05/2022 12:35, John Ogness wrote:
> The uart_ops startup() callback is called without interrupts
> disabled and without port->lock locked, relatively late during the
> boot process (from the call path of console_on_rootfs()). If the
> device is a console, it was already previously registered and could
> be actively printing messages.
> 
> Since the startup() callback is reading/writing registers used by
> the console write() callback (AML_UART_CONTROL), its access must
> be synchronized using the port->lock. Currently it is not.
> 
> The startup() callback is the only function that explicitly enables
> interrupts. Without the synchronization, it is possible that
> interrupts become accidentally permanently disabled.
> 
> CPU0                           CPU1
> meson_serial_console_write     meson_uart_startup
> --------------------------     ------------------
> spin_lock(port->lock)
> val = readl(AML_UART_CONTROL)
> uart_console_write()
>                                 writel(INT_EN, AML_UART_CONTROL)
> writel(val, AML_UART_CONTROL)
> spin_unlock(port->lock)
> 
> Add port->lock synchronization to meson_uart_startup() to avoid
> racing with meson_serial_console_write().
> 
> Also add detailed comments to meson_uart_reset() explaining why it
> is *not* using port->lock synchronization.
> 
> Link: https://lore.kernel.org/lkml/2a82eae7-a256-f70c-fd82-4e510750906e@samsung.com
> Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
> Signed-off-by: John Ogness <john.ogness@linutronix.de>
> Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
> ---
>   drivers/tty/serial/meson_uart.c | 13 +++++++++++++
>   1 file changed, 13 insertions(+)
> 
> diff --git a/drivers/tty/serial/meson_uart.c b/drivers/tty/serial/meson_uart.c
> index 2bf1c57e0981..39021dac09cc 100644
> --- a/drivers/tty/serial/meson_uart.c
> +++ b/drivers/tty/serial/meson_uart.c
> @@ -253,6 +253,14 @@ static const char *meson_uart_type(struct uart_port *port)
>   	return (port->type == PORT_MESON) ? "meson_uart" : NULL;
>   }
>   
> +/*
> + * This function is called only from probe() using a temporary io mapping
> + * in order to perform a reset before setting up the device. Since the
> + * temporarily mapped region was successfully requested, there can be no
> + * console on this port at this time. Hence it is not necessary for this
> + * function to acquire the port->lock. (Since there is no console on this
> + * port at this time, the port->lock is not initialized yet.)
> + */
>   static void meson_uart_reset(struct uart_port *port)
>   {
>   	u32 val;
> @@ -267,9 +275,12 @@ static void meson_uart_reset(struct uart_port *port)
>   
>   static int meson_uart_startup(struct uart_port *port)
>   {
> +	unsigned long flags;
>   	u32 val;
>   	int ret = 0;
>   
> +	spin_lock_irqsave(&port->lock, flags);
> +
>   	val = readl(port->membase + AML_UART_CONTROL);
>   	val |= AML_UART_CLEAR_ERR;
>   	writel(val, port->membase + AML_UART_CONTROL);
> @@ -285,6 +296,8 @@ static int meson_uart_startup(struct uart_port *port)
>   	val = (AML_UART_RECV_IRQ(1) | AML_UART_XMIT_IRQ(port->fifosize / 2));
>   	writel(val, port->membase + AML_UART_MISC);
>   
> +	spin_unlock_irqrestore(&port->lock, flags);
> +
>   	ret = request_irq(port->irq, meson_uart_interrupt, 0,
>   			  port->name, port);
>   
> 
> base-commit: 672c0c5173427e6b3e2a9bbb7be51ceeec78093a

Thanks for fixing this, it may also fix an uart apparent lockup I encountered
several time while developing on the platform, but the target was still alive
so it matches.

So I'll add:
Fixes: ff7693d079e5 ("ARM: meson: serial: add MesonX SoC on-chip uart driver")

and

Acked-by: Neil Armstrong <narmstrong@baylibre.com>