[PATCH] x86, setup: Check early serial console per string instead of one char

[PATCH] x86, setup: Check early serial console per string instead of one char

[Yinghai Lu]

Move out serial_putchar() calling out of putchar
Let puts() to call serial_putchar() directly.

So only need to check early_serial_base per string.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>

---
 arch/x86/boot/tty.c |   14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

Index: linux-2.6/arch/x86/boot/tty.c
===================================================================
--- linux-2.6.orig/arch/x86/boot/tty.c
+++ linux-2.6/arch/x86/boot/tty.c
@@ -52,16 +52,22 @@ static void __attribute__((section(".ini
 void __attribute__((section(".inittext"))) putchar(int ch)
 {
 	if (ch == '\n')
-		putchar('\r');	/* \n -> \r\n */
+		bios_putchar('\r');	/* \n -> \r\n */
 
 	bios_putchar(ch);
-
-	if (early_serial_base != 0)
-		serial_putchar(ch);
 }
 
 void __attribute__((section(".inittext"))) puts(const char *str)
 {
+	if (early_serial_base) {
+		const char *s = str;
+		while (*s) {
+			if (*s == '\n')
+				serial_putchar('\r');
+			serial_putchar(*s++);
+		}
+	}
+
 	while (*str)
 		putchar(*str++);
 }

Re: [PATCH] x86, setup: Check early serial console per string instead of one char

[H. Peter Anvin]

On 10/11/2010 10:50 AM, Yinghai Lu wrote:
> 
> Move out serial_putchar() calling out of putchar
> Let puts() to call serial_putchar() directly.
> 
> So only need to check early_serial_base per string.
> 
> Signed-off-by: Yinghai Lu <yinghai@kernel.org>
> 

This seems to add a bunch of code for zero benefit.  Testing a global
variable is zero cost compared to either a BIOS call or poking a serial
port.

	-hpa

Re: [PATCH] x86, setup: Check early serial console per string instead of one char

[Yinghai Lu]

On 10/11/2010 11:19 AM, H. Peter Anvin wrote:
> On 10/11/2010 10:50 AM, Yinghai Lu wrote:
>>
>> Move out serial_putchar() calling out of putchar
>> Let puts() to call serial_putchar() directly.
>>
>> So only need to check early_serial_base per string.
>>
>> Signed-off-by: Yinghai Lu <yinghai@kernel.org>
>>
> 
> This seems to add a bunch of code for zero benefit.  Testing a global
> variable is zero cost compared to either a BIOS call or poking a serial
> port.

some server have bios serial console redirection support, and bios int call could send char to screen and serial port.

in that case, take out serial_putchar from putchar could produce more readable print out from serial console

Thanks

Yinghai

Re: [PATCH] x86, setup: Check early serial console per string instead of one char

[H. Peter Anvin]

On 10/11/2010 12:47 PM, Yinghai Lu wrote:
>>
>> This seems to add a bunch of code for zero benefit.  Testing a global
>> variable is zero cost compared to either a BIOS call or poking a serial
>> port.
> 
> some server have bios serial console redirection support, and bios int call could send char to screen and serial port.
> 
> in that case, take out serial_putchar from putchar could produce more readable print out from serial console
> 

That doesn't appear to have anything to do with this code change.  What
you're saying is that "if BIOS does serial console redirection, we
shouldn't output to both the serial console and the BIOS console", which
is of course a truism, but that doesn't justify in any way this
particular code reconfiguration.

	-hpa