IP32: serial console broken with current git

IP32: serial console broken with current git

[Martin Michlmayr]

With current git (and 2.6.23-rc1), nothing shows up on the serial
console on IP32.  Ricardo Mendoza commented on this on IRC:

> I think it was that using iobase member in the plat_serial8250_port
> struct was not working, swapping to membase gave console messages
> but kind of stopped printing messages at some point (further in than
> first line of C tho)

He also sent me a patch to test.  With this patch, I get serial
console output - but only until:

| Serial: 8250/16550 driver $Revision: 1.90 $ 4 ports, IRQ sharing disabled
| serial8250.0: ttyS0 at MMIO 0x0 (irq = 53) is a 16550A
| console [ttyS0] enabled

Maybe Ricardo can post his patch for comments and someone can look
into the second issue.
-- 
Martin Michlmayr
http://www.cyrius.com/

Re: IP32: serial console broken with current git

[Ricardo Mendoza]

[-- Attachment #1: Type: text/plain, Size: 1259 bytes --]

Martin Michlmayr wrote:
> With current git (and 2.6.23-rc1), nothing shows up on the serial
> console on IP32.  Ricardo Mendoza commented on this on IRC:
> 
>> I think it was that using iobase member in the plat_serial8250_port
>> struct was not working, swapping to membase gave console messages
>> but kind of stopped printing messages at some point (further in than
>> first line of C tho)
> 
> He also sent me a patch to test.  With this patch, I get serial
> console output - but only until:
> 
> | Serial: 8250/16550 driver $Revision: 1.90 $ 4 ports, IRQ sharing disabled
> | serial8250.0: ttyS0 at MMIO 0x0 (irq = 53) is a 16550A
> | console [ttyS0] enabled
> 
> Maybe Ricardo can post his patch for comments and someone can look
> into the second issue.


The patch is the following, please ignore the rewrite of the MACE_PORT
specifications, those are not needed (I don't remember why I did that at
all).

As Martin said, the output stops on the 'console [ttyS0] enabled'
message. I don't know much about the serial code internals but I presume
something is getting trashed on the console setup.. or perhaps it's just
something really dumb like a missing call or something. Someone
enlightened in the issue can perhaps shed a light.


     Ricardo



[-- Attachment #2: plat_dev.patch --]
[-- Type: text/x-patch, Size: 1722 bytes --]

diff --git a/arch/mips/sgi-ip32/ip32-platform.c b/arch/mips/sgi-ip32/ip32-platform.c
index ba3697e..6ad2ecc 100644
--- a/arch/mips/sgi-ip32/ip32-platform.c
+++ b/arch/mips/sgi-ip32/ip32-platform.c
@@ -10,25 +10,29 @@
 #include <linux/platform_device.h>
 #include <linux/serial_8250.h>
 
+#include <asm/addrspace.h>
 #include <asm/ip32/mace.h>
 #include <asm/ip32/ip32_ints.h>
 
-/*
- * .iobase isn't a constant (in the sense of C) so we fill it in at runtime.
- */
-#define MACE_PORT(int)							\
-{									\
-	.irq		= int,						\
-	.uartclk	= 1843200,					\
-	.iotype		= UPIO_MEM,					\
-	.flags		= UPF_SKIP_TEST,				\
-	.regshift	= 8,						\
-}
 
 static struct plat_serial8250_port uart8250_data[] = {
-	MACE_PORT(MACEISA_SERIAL1_IRQ),
-	MACE_PORT(MACEISA_SERIAL2_IRQ),
-	{ },
+	[0] = {
+		.irq            = MACEISA_SERIAL1_IRQ,
+	        .uartclk        = 1843200,
+        	.iotype         = UPIO_MEM,
+	        .flags          = UPF_SKIP_TEST,
+        	.regshift       = 8,
+	},
+	[1] = {
+                .irq            = MACEISA_SERIAL2_IRQ,
+                .uartclk        = 1843200,
+                .iotype         = UPIO_MEM,
+                .flags          = UPF_SKIP_TEST,
+                .regshift       = 8,
+	},
+	{
+		0,
+	},
 };
 
 static struct platform_device uart8250_device = {
@@ -41,8 +45,8 @@ static struct platform_device uart8250_device = {
 
 static int __init uart8250_init(void)
 {
-	uart8250_data[0].iobase = (unsigned long) &mace->isa.serial1;
-	uart8250_data[1].iobase = (unsigned long) &mace->isa.serial1;
+	uart8250_data[0].membase = (char *) &mace->isa.serial1;
+	uart8250_data[1].membase = (char *) &mace->isa.serial2;
 
 	return platform_device_register(&uart8250_device);
 }

Re: IP32: serial console broken with current git

[Ralf Baechle]

On Tue, Sep 25, 2007 at 12:25:54PM -0400, Ricardo Mendoza wrote:

> Martin Michlmayr wrote:
> > With current git (and 2.6.23-rc1), nothing shows up on the serial
> > console on IP32.  Ricardo Mendoza commented on this on IRC:
> > 
> >> I think it was that using iobase member in the plat_serial8250_port
> >> struct was not working, swapping to membase gave console messages
> >> but kind of stopped printing messages at some point (further in than
> >> first line of C tho)
> > 
> > He also sent me a patch to test.  With this patch, I get serial
> > console output - but only until:
> > 
> > | Serial: 8250/16550 driver $Revision: 1.90 $ 4 ports, IRQ sharing disabled
> > | serial8250.0: ttyS0 at MMIO 0x0 (irq = 53) is a 16550A
> > | console [ttyS0] enabled
> > 
> > Maybe Ricardo can post his patch for comments and someone can look
> > into the second issue.
> 
> 
> The patch is the following, please ignore the rewrite of the MACE_PORT
> specifications, those are not needed (I don't remember why I did that at
> all).
> 
> As Martin said, the output stops on the 'console [ttyS0] enabled'
> message. I don't know much about the serial code internals but I presume
> something is getting trashed on the console setup.. or perhaps it's just
> something really dumb like a missing call or something. Someone
> enlightened in the issue can perhaps shed a light.

The membase part of it looks right.

  Ralf