From mboxrd@z Thu Jan  1 00:00:00 1970
From: Peter Hurley <peter@hurleysoftware.com>
Subject: Re: [PATCH] tty/serial: atmel: use proper accessors
Date: Mon, 20 Jul 2015 15:05:39 -0400
Message-ID: <55AD4683.3090708@hurleysoftware.com>
References: <1437400572-16575-1-git-send-email-alexandre.belloni@free-electrons.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <1437400572-16575-1-git-send-email-alexandre.belloni@free-electrons.com>
Sender: linux-kernel-owner@vger.kernel.org
To: Alexandre Belloni <alexandre.belloni@free-electrons.com>, Nicolas Ferre <nicolas.ferre@atmel.com>, Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.com>, linux-serial@vger.kernel.org, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, Cyrille Pitchen <cyrille.pitchen@atmel.com>
List-Id: linux-serial@vger.kernel.org

Hi Alexandre,

On 07/20/2015 09:56 AM, Alexandre Belloni wrote:
> On ARM, it is necessary to use the _relaxed versions of read[bl] and
> write[bl] to ensure we do the correct endianness conversion when running in
> big endian.
> Keep using avr32 __raw version on avr32 to avoid breakage.
> 
> Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
> ---
> 
> This patch applies after [PATCH v4 0/5] tty/serial: at91: add support to FIFOs
> http://lists.infradead.org/pipermail/linux-arm-kernel/2015-July/353704.html
> 
>  drivers/tty/serial/atmel_serial.c | 22 ++++++++++++++++++++++
>  1 file changed, 22 insertions(+)
> 
> diff --git a/drivers/tty/serial/atmel_serial.c b/drivers/tty/serial/atmel_serial.c
> index e91b3b2f0590..451acf054281 100644
> --- a/drivers/tty/serial/atmel_serial.c
> +++ b/drivers/tty/serial/atmel_serial.c
> @@ -194,6 +194,7 @@ to_atmel_uart_port(struct uart_port *uart)
>  	return container_of(uart, struct atmel_uart_port, uart);
>  }
>  
> +#ifdef CONFIG_AVR32

The whole point of i/o accessors is to encapsulate arch-specific
differences, so one of these arches isn't doing it right, or the wrong
i/o accessors are being used.

In any event, serial drivers can't start taking on per-arch conditional
compilation.

Regards,
Peter Hurley

PS - The $subject doesn't really match the patch; this is basically adding
big-endian support on little-endian arches.

>  static inline u32 atmel_uart_readl(struct uart_port *port, u32 reg)
>  {
>  	return __raw_readl(port->membase + reg);
> @@ -213,6 +214,27 @@ static inline void atmel_uart_writeb(struct uart_port *port, u32 reg, u8 value)
>  {
>  	__raw_writeb(value, port->membase + reg);
>  }
> +#else
> +static inline u32 atmel_uart_readl(struct uart_port *port, u32 reg)
> +{
> +	return readl_relaxed(port->membase + reg);
> +}
> +
> +static inline void atmel_uart_writel(struct uart_port *port, u32 reg, u32 value)
> +{
> +	writel_relaxed(value, port->membase + reg);
> +}
> +
> +static inline u8 atmel_uart_readb(struct uart_port *port, u32 reg)
> +{
> +	return readb_relaxed(port->membase + reg);
> +}
> +
> +static inline void atmel_uart_writeb(struct uart_port *port, u32 reg, u8 value)
> +{
> +	writeb_relaxed(value, port->membase + reg);
> +}
> +#endif
>  
>  #ifdef CONFIG_SERIAL_ATMEL_PDC
>  static bool atmel_use_pdc_rx(struct uart_port *port)
>