[PATCH V2 4/9] arch: mips: ralink: add tty detection

[PATCH V2 4/9] arch: mips: ralink: add tty detection

[John Crispin]

MT7688 has several uarts that can be used for console. There are several
boards in the wild, that use ttyS1 or ttyS2. This patch applies a simply
autodetection routine to figure out which ttyS the bootloader used as
console. The uarts come up in 6 bit mode by default. The bootloader will
have set 8 bit mode on the console. Find that 8bit tty and use it.

Signed-off-by: John Crispin <blogic@openwrt.org>
---
Changes in V2:
* remove superflous inline definition

 arch/mips/ralink/early_printk.c |   26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/arch/mips/ralink/early_printk.c b/arch/mips/ralink/early_printk.c
index 255d695..3c59ffe 100644
--- a/arch/mips/ralink/early_printk.c
+++ b/arch/mips/ralink/early_printk.c
@@ -25,11 +25,13 @@
 #define MT7628_CHIP_NAME1	0x20203832
 
 #define UART_REG_TX		0x04
+#define UART_REG_LCR		0x0c
 #define UART_REG_LSR		0x14
 #define UART_REG_LSR_RT2880	0x1c
 
 static __iomem void *uart_membase = (__iomem void *) KSEG1ADDR(EARLY_UART_BASE);
 static __iomem void *chipid_membase = (__iomem void *) KSEG1ADDR(CHIPID_BASE);
+static int init_complete;
 
 static inline void uart_w32(u32 val, unsigned reg)
 {
@@ -47,8 +49,32 @@ static inline int soc_is_mt7628(void)
 		(__raw_readl(chipid_membase) == MT7628_CHIP_NAME1);
 }
 
+static void find_uart_base(void)
+{
+	int i;
+
+	if (!soc_is_mt7628())
+		return;
+
+	for (i = 0; i < 3; i++) {
+		u32 reg = uart_r32(UART_REG_LCR + (0x100 * i));
+
+		if (!reg)
+			continue;
+
+		uart_membase = (__iomem void *) KSEG1ADDR(EARLY_UART_BASE +
+							  (0x100 * i));
+		break;
+	}
+}
+
 void prom_putchar(unsigned char ch)
 {
+	if (!init_complete) {
+		find_uart_base();
+		init_complete = 1;
+	}
+
 	if (IS_ENABLED(CONFIG_SOC_MT7621) || soc_is_mt7628()) {
 		uart_w32(ch, UART_TX);
 		while ((uart_r32(UART_REG_LSR) & UART_LSR_THRE) == 0)
-- 
1.7.10.4

Re: [PATCH V2 4/9] arch: mips: ralink: add tty detection

[Sergei Shtylyov]

Hello.

On 11/5/2015 5:59 AM, John Crispin wrote:

> MT7688 has several uarts that can be used for console. There are several
> boards in the wild, that use ttyS1 or ttyS2. This patch applies a simply
> autodetection routine to figure out which ttyS the bootloader used as
> console. The uarts come up in 6 bit mode by default. The bootloader will
> have set 8 bit mode on the console. Find that 8bit tty and use it.
>
> Signed-off-by: John Crispin <blogic@openwrt.org>
> ---
> Changes in V2:
> * remove superflous inline definition
>
>   arch/mips/ralink/early_printk.c |   26 ++++++++++++++++++++++++++
>   1 file changed, 26 insertions(+)
>
> diff --git a/arch/mips/ralink/early_printk.c b/arch/mips/ralink/early_printk.c
> index 255d695..3c59ffe 100644
> --- a/arch/mips/ralink/early_printk.c
> +++ b/arch/mips/ralink/early_printk.c
[...]
> @@ -47,8 +49,32 @@ static inline int soc_is_mt7628(void)
>   		(__raw_readl(chipid_membase) == MT7628_CHIP_NAME1);
>   }
>
> +static void find_uart_base(void)
> +{
> +	int i;
> +
> +	if (!soc_is_mt7628())
> +		return;
> +
> +	for (i = 0; i < 3; i++) {
> +		u32 reg = uart_r32(UART_REG_LCR + (0x100 * i));

    Inner parens not needed, the operator precedence is natural.

> +
> +		if (!reg)
> +			continue;
> +
> +		uart_membase = (__iomem void *) KSEG1ADDR(EARLY_UART_BASE +
> +							  (0x100 * i));

    Likewise.
    Sorry for not noticing this before.

MBR, Sergei

Re: [PATCH V2 4/9] arch: mips: ralink: add tty detection

[John Crispin]

Hi,

On 05/11/2015 15:07, Sergei Shtylyov wrote:
> Hello.
> 
> On 11/5/2015 5:59 AM, John Crispin wrote:
> 
>> MT7688 has several uarts that can be used for console. There are several
>> boards in the wild, that use ttyS1 or ttyS2. This patch applies a simply
>> autodetection routine to figure out which ttyS the bootloader used as
>> console. The uarts come up in 6 bit mode by default. The bootloader will
>> have set 8 bit mode on the console. Find that 8bit tty and use it.
>>
>> Signed-off-by: John Crispin <blogic@openwrt.org>
>> ---
>> Changes in V2:
>> * remove superflous inline definition
>>
>>   arch/mips/ralink/early_printk.c |   26 ++++++++++++++++++++++++++
>>   1 file changed, 26 insertions(+)
>>
>> diff --git a/arch/mips/ralink/early_printk.c
>> b/arch/mips/ralink/early_printk.c
>> index 255d695..3c59ffe 100644
>> --- a/arch/mips/ralink/early_printk.c
>> +++ b/arch/mips/ralink/early_printk.c
> [...]
>> @@ -47,8 +49,32 @@ static inline int soc_is_mt7628(void)
>>           (__raw_readl(chipid_membase) == MT7628_CHIP_NAME1);
>>   }
>>
>> +static void find_uart_base(void)
>> +{
>> +    int i;
>> +
>> +    if (!soc_is_mt7628())
>> +        return;
>> +
>> +    for (i = 0; i < 3; i++) {
>> +        u32 reg = uart_r32(UART_REG_LCR + (0x100 * i));
> 
>    Inner parens not needed, the operator precedence is natural.
> 

"not needed" means "should be removed" or "not needed".

checkpatch.pl certainly did not complain and a quick look around
instantly yielded lots of places in the kernel where this is done. imho
the brackets make it more readable

	John

>> +
>> +        if (!reg)
>> +            continue;
>> +
>> +        uart_membase = (__iomem void *) KSEG1ADDR(EARLY_UART_BASE +
>> +                              (0x100 * i));
> 
>    Likewise.
>    Sorry for not noticing this before.
> 
> MBR, Sergei
> 
> 

Re: [PATCH V2 4/9] arch: mips: ralink: add tty detection

[Sergei Shtylyov]

On 11/05/2015 05:27 PM, John Crispin wrote:

>>> MT7688 has several uarts that can be used for console. There are several
>>> boards in the wild, that use ttyS1 or ttyS2. This patch applies a simply
>>> autodetection routine to figure out which ttyS the bootloader used as
>>> console. The uarts come up in 6 bit mode by default. The bootloader will
>>> have set 8 bit mode on the console. Find that 8bit tty and use it.
>>>
>>> Signed-off-by: John Crispin <blogic@openwrt.org>
>>> ---
>>> Changes in V2:
>>> * remove superflous inline definition
>>>
>>>    arch/mips/ralink/early_printk.c |   26 ++++++++++++++++++++++++++
>>>    1 file changed, 26 insertions(+)
>>>
>>> diff --git a/arch/mips/ralink/early_printk.c
>>> b/arch/mips/ralink/early_printk.c
>>> index 255d695..3c59ffe 100644
>>> --- a/arch/mips/ralink/early_printk.c
>>> +++ b/arch/mips/ralink/early_printk.c
>> [...]
>>> @@ -47,8 +49,32 @@ static inline int soc_is_mt7628(void)
>>>            (__raw_readl(chipid_membase) == MT7628_CHIP_NAME1);
>>>    }
>>>
>>> +static void find_uart_base(void)
>>> +{
>>> +    int i;
>>> +
>>> +    if (!soc_is_mt7628())
>>> +        return;
>>> +
>>> +    for (i = 0; i < 3; i++) {
>>> +        u32 reg = uart_r32(UART_REG_LCR + (0x100 * i));
>>
>>     Inner parens not needed, the operator precedence is natural.
>>
>
> "not needed" means "should be removed" or "not needed".

    That's completely up to you.

> checkpatch.pl certainly did not complain and a quick look around
> instantly yielded lots of places in the kernel where this is done. imho
> the brackets make it more readable

    Do you really always write 2 + (3 * 4) when you e.g. use scientific 
calculator?

[...]

MBR, Sergei

Re: [PATCH V2 4/9] arch: mips: ralink: add tty detection

[John Crispin]

On 05/11/2015 18:21, Sergei Shtylyov wrote:
> On 11/05/2015 05:27 PM, John Crispin wrote:
> 
>>>> MT7688 has several uarts that can be used for console. There are
>>>> several
>>>> boards in the wild, that use ttyS1 or ttyS2. This patch applies a
>>>> simply
>>>> autodetection routine to figure out which ttyS the bootloader used as
>>>> console. The uarts come up in 6 bit mode by default. The bootloader
>>>> will
>>>> have set 8 bit mode on the console. Find that 8bit tty and use it.
>>>>
>>>> Signed-off-by: John Crispin <blogic@openwrt.org>
>>>> ---
>>>> Changes in V2:
>>>> * remove superflous inline definition
>>>>
>>>>    arch/mips/ralink/early_printk.c |   26 ++++++++++++++++++++++++++
>>>>    1 file changed, 26 insertions(+)
>>>>
>>>> diff --git a/arch/mips/ralink/early_printk.c
>>>> b/arch/mips/ralink/early_printk.c
>>>> index 255d695..3c59ffe 100644
>>>> --- a/arch/mips/ralink/early_printk.c
>>>> +++ b/arch/mips/ralink/early_printk.c
>>> [...]
>>>> @@ -47,8 +49,32 @@ static inline int soc_is_mt7628(void)
>>>>            (__raw_readl(chipid_membase) == MT7628_CHIP_NAME1);
>>>>    }
>>>>
>>>> +static void find_uart_base(void)
>>>> +{
>>>> +    int i;
>>>> +
>>>> +    if (!soc_is_mt7628())
>>>> +        return;
>>>> +
>>>> +    for (i = 0; i < 3; i++) {
>>>> +        u32 reg = uart_r32(UART_REG_LCR + (0x100 * i));
>>>
>>>     Inner parens not needed, the operator precedence is natural.
>>>
>>
>> "not needed" means "should be removed" or "not needed".
> 
>    That's completely up to you.
> 
>> checkpatch.pl certainly did not complain and a quick look around
>> instantly yielded lots of places in the kernel where this is done. imho
>> the brackets make it more readable
> 
>    Do you really always write 2 + (3 * 4) when you e.g. use scientific
> calculator?

no that would be silly. when using a calculator i know what i am typing.
unless there is someone looking over my shoulder i am the only one that
needs to understand.

when i write software that others are supposed to read which has
"register + (bank * bank offset)" constructs in it then i tend to add
brackets.

	John

> 
> [...]
> 
> MBR, Sergei