* [PATCH v1 2/2] atmel_serial: set 32-bit access by default
[not found] ` <1438261591-13130-1-git-send-email-andriy.shevchenko-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
@ 2015-07-30 13:06 ` Andy Shevchenko
2015-07-30 13:36 ` [PATCH v1 1/2] dt: atmel-usart: document new I/O data register width property Nicolas Ferre
1 sibling, 0 replies; 4+ messages in thread
From: Andy Shevchenko @ 2015-07-30 13:06 UTC (permalink / raw)
To: Rob Herring, Nicolas Ferre, Greg Kroah-Hartman,
linux-serial-u79uwXL29TY76Z2rM5mHXA,
devicetree-u79uwXL29TY76Z2rM5mHXA, Cyrille Pitchen,
Hans-Christian Egtvedt,
alexandre.belloni-wi1+55ScJUtKEb57/3fJTNBPR1lH4CV8
Cc: Andy Shevchenko
The commit b5199d468177 (tty/serial: at91: add support to FIFOs) changed data
stream 32-bit accesses in the driver to 8-bit. This, unfortunately, breaks data
register access on ATNGW100, where the IP needs data register accesses to be
long accesses (all other accesses appear to be OK).
This change introduces a new master variable to allow interface drivers to
specify that 8-bit data transfer I/O is required. This change also introduces
the ability to set this variable via device tree bindings in the driver.
Fixes: b5199d468177 (tty/serial: at91: add support to FIFOs)
Signed-off-by: Andy Shevchenko <andriy.shevchenko-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
---
drivers/tty/serial/atmel_serial.c | 45 +++++++++++++++++++++++++++++++++------
1 file changed, 39 insertions(+), 6 deletions(-)
diff --git a/drivers/tty/serial/atmel_serial.c b/drivers/tty/serial/atmel_serial.c
index e91b3b2f..0312bce 100644
--- a/drivers/tty/serial/atmel_serial.c
+++ b/drivers/tty/serial/atmel_serial.c
@@ -155,6 +155,7 @@ struct atmel_uart_port {
u32 rts_high;
u32 rts_low;
bool ms_irq_enabled;
+ u32 reg_io_width; /* I/O width in bytes */
bool is_usart; /* usart or uart */
struct timer_list uart_timer; /* uart timer */
@@ -214,6 +215,34 @@ static inline void atmel_uart_writeb(struct uart_port *port, u32 reg, u8 value)
__raw_writeb(value, port->membase + reg);
}
+static inline u8 atmel_uart_read_reg(struct uart_port *port, u32 reg)
+{
+ struct atmel_uart_port *atmel_port = to_atmel_uart_port(port);
+
+ switch (atmel_port->reg_io_width) {
+ case 1:
+ return atmel_uart_readb(port, reg);
+ case 4:
+ default:
+ return atmel_uart_readl(port, reg);
+ }
+}
+
+static inline void atmel_uart_write_reg(struct uart_port *port, u32 reg, u8 value)
+{
+ struct atmel_uart_port *atmel_port = to_atmel_uart_port(port);
+
+ switch (atmel_port->reg_io_width) {
+ case 1:
+ atmel_uart_writeb(port, reg, value);
+ break;
+ case 4:
+ default:
+ atmel_uart_writel(port, reg, value);
+ break;
+ }
+}
+
#ifdef CONFIG_SERIAL_ATMEL_PDC
static bool atmel_use_pdc_rx(struct uart_port *port)
{
@@ -658,7 +687,7 @@ static void atmel_rx_chars(struct uart_port *port)
status = atmel_uart_readl(port, ATMEL_US_CSR);
while (status & ATMEL_US_RXRDY) {
- ch = atmel_uart_readb(port, ATMEL_US_RHR);
+ ch = atmel_uart_read_reg(port, ATMEL_US_RHR);
/*
* note that the error handling code is
@@ -709,7 +738,7 @@ static void atmel_tx_chars(struct uart_port *port)
if (port->x_char &&
(atmel_uart_readl(port, ATMEL_US_CSR) & atmel_port->tx_done_mask)) {
- atmel_uart_writeb(port, ATMEL_US_THR, port->x_char);
+ atmel_uart_write_reg(port, ATMEL_US_THR, port->x_char);
port->icount.tx++;
port->x_char = 0;
}
@@ -718,7 +747,7 @@ static void atmel_tx_chars(struct uart_port *port)
while (atmel_uart_readl(port, ATMEL_US_CSR) &
atmel_port->tx_done_mask) {
- atmel_uart_writeb(port, ATMEL_US_THR, xmit->buf[xmit->tail]);
+ atmel_uart_write_reg(port, ATMEL_US_THR, xmit->buf[xmit->tail]);
xmit->tail = (xmit->tail + 1) & (UART_XMIT_SIZE - 1);
port->icount.tx++;
if (uart_circ_empty(xmit))
@@ -2294,7 +2323,7 @@ static int atmel_poll_get_char(struct uart_port *port)
while (!(atmel_uart_readl(port, ATMEL_US_CSR) & ATMEL_US_RXRDY))
cpu_relax();
- return atmel_uart_readb(port, ATMEL_US_RHR);
+ return atmel_uart_read_reg(port, ATMEL_US_RHR);
}
static void atmel_poll_put_char(struct uart_port *port, unsigned char ch)
@@ -2302,7 +2331,7 @@ static void atmel_poll_put_char(struct uart_port *port, unsigned char ch)
while (!(atmel_uart_readl(port, ATMEL_US_CSR) & ATMEL_US_TXRDY))
cpu_relax();
- atmel_uart_writeb(port, ATMEL_US_THR, ch);
+ atmel_uart_write_reg(port, ATMEL_US_THR, ch);
}
#endif
@@ -2409,7 +2438,7 @@ static void atmel_console_putchar(struct uart_port *port, int ch)
{
while (!(atmel_uart_readl(port, ATMEL_US_CSR) & ATMEL_US_TXRDY))
cpu_relax();
- atmel_uart_writeb(port, ATMEL_US_THR, ch);
+ atmel_uart_write_reg(port, ATMEL_US_THR, ch);
}
/*
@@ -2762,6 +2791,10 @@ static int atmel_serial_probe(struct platform_device *pdev)
port = &atmel_ports[ret];
port->backup_imr = 0;
port->uart.line = ret;
+
+ of_property_read_u32(pdev->dev.of_node, "reg-io-width",
+ &port->reg_io_width);
+
atmel_serial_probe_fifos(port, pdev);
spin_lock_init(&port->lock_suspended);
--
2.5.0
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 4+ messages in thread* Re: [PATCH v1 1/2] dt: atmel-usart: document new I/O data register width property
[not found] ` <1438261591-13130-1-git-send-email-andriy.shevchenko-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
2015-07-30 13:06 ` [PATCH v1 2/2] atmel_serial: set 32-bit access by default Andy Shevchenko
@ 2015-07-30 13:36 ` Nicolas Ferre
[not found] ` <55BA2849.2050803-AIFe0yeh4nAAvxtiuMwx3w@public.gmane.org>
1 sibling, 1 reply; 4+ messages in thread
From: Nicolas Ferre @ 2015-07-30 13:36 UTC (permalink / raw)
To: Andy Shevchenko, Rob Herring, Greg Kroah-Hartman,
linux-serial-u79uwXL29TY76Z2rM5mHXA, Cyrille Pitchen,
Hans-Christian Egtvedt
Cc: devicetree-u79uwXL29TY76Z2rM5mHXA,
alexandre.belloni-wi1+55ScJUtKEb57/3fJTNBPR1lH4CV8,
linux-arm-kernel
Le 30/07/2015 15:06, Andy Shevchenko a écrit :
> This change documents a new property for the Atmel serial device, allowing an
> implementer to specify either four bytes or one byte access to the controller
> data register.
>
> This supports a change that unbreaks this driver on ATNGW100 board.
>
> Signed-off-by: Andy Shevchenko <andriy.shevchenko-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
> ---
> Documentation/devicetree/bindings/serial/atmel-usart.txt | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/Documentation/devicetree/bindings/serial/atmel-usart.txt b/Documentation/devicetree/bindings/serial/atmel-usart.txt
> index e6e6142..a8c290a7 100644
> --- a/Documentation/devicetree/bindings/serial/atmel-usart.txt
> +++ b/Documentation/devicetree/bindings/serial/atmel-usart.txt
> @@ -6,6 +6,8 @@ Required properties:
> additional mode or an USART new feature.
> For the dbgu UART, use "atmel,<chip>-dbgu", "atmel,<chip>-usart"
> - reg: Should contain registers location and length
> +- reg-io-width: The I/O register width (in bytes) implemented by
> + this device. Supported values are 1 or 4 (the default).
Well, sorry but no.
The proper way to implement this is to cling to certain compatible
strings. But I'm pretty sure that this issue shall not be fixed like this.
But I think that Cyrille is about to propose another version. Let's wait
for his code...
Bye,
> - interrupts: Should contain interrupt
> - clock-names: tuple listing input clock names.
> Required elements: "usart"
>
--
Nicolas Ferre
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 4+ messages in thread