early printk for x86

I copied the x86_64 early printk support for plain x86. Is anyone opposed to me sending this on to Linus?

Dave Hansen
[email protected]

diff -urN linux-2.5.47-clean/arch/i386/Kconfig linux-2.5.47/arch/i386/Kconfig
@@ -1574,6 +1574,26 @@
 	  that can still be used by old drivers that are being ported from
 	  2.0/2.2.
 
+config EARLY_PRINTK
+	bool "Early console support"
+	default n
+	depends on DEBUG_KERNEL
+	help
+	  Write kernel log output directly into the VGA buffer or serial port. 
+	  This is useful for kernel debugging when your machine crashes very 
+	  early before the console code is initialized. For normal operation 
+	  it is not recommended because it looks ugly and doesn't cooperate 
+	  with klogd/syslogd or the X server.You should normally N here, 
+	  unless you want to debug such a crash.
+
+	  Syntax: earlyprintk=vga
+		  earlyprintk=serial[,ttySn[,baudrate]] 
+	  Append ,keep to not disable it when the real console takes over.
+	  Only vga or serial at a time, not both.
+	  Currently only ttyS0 and ttyS1 are supported. 
+	  Interaction with the standard serial driver is not very good. 
+	  The VGA output is eventually overwritten by the real console.
+
 config MAGIC_SYSRQ
 	bool "Magic SysRq key"
 	depends on DEBUG_KERNEL

diff -urN linux-2.5.47-clean/arch/i386/kernel/Makefile linux-2.5.47/arch/i386/kernel/Makefile
@@ -28,6 +28,7 @@
 obj-$(CONFIG_X86_NUMAQ)		+= numaq.o
 obj-$(CONFIG_PROFILING)		+= profile.o
 obj-$(CONFIG_EDD)             	+= edd.o
+obj-$(CONFIG_EARLY_PRINTK)	+= early_printk.o
 
 EXTRA_AFLAGS   := -traditional
 

diff -urN linux-2.5.47-clean/arch/i386/kernel/setup.c linux-2.5.47/arch/i386/kernel/setup.c
@@ -84,6 +84,7 @@
 
 extern void early_cpu_init(void);
 extern void dmi_scan_machine(void);
+extern void setup_early_printk(char *);
 extern int root_mountflags;
 extern char _text, _etext, _edata, _end;
 extern int blk_nohighio;

@@ -836,7 +837,12 @@
 void __init setup_arch(char **cmdline_p)
 {
 	unsigned long max_low_pfn;
+	char* s;
 
+	s = strstr(saved_command_line, "earlyprintk=");
+	if (s != NULL)
+		setup_early_printk(s+12);
+	
 	pre_setup_arch_hook();
 	early_cpu_init();
 

No problem from my side.

although it may make sense to just put in the #ifdef __i386__ for vgabase and then #include it.

-Andi

VGA and serial are certainly not hammer+ia32-specific. Make the generic parts generic... the arch-specific components would need to change early-foo base addresses perhaps, but otherwise, it's pretty generic.

Jeff

VGA and serial are certainly not hammer+ia32-specific. Make the generic parts generic... the arch-specific components would need to change early-foo base addresses perhaps, but otherwise, it's pretty generic.

Jeff

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Take 2.

- Move the x86_64 early_printk.c into kernel/
- move some of the basic defines into linux/early_printk.h and asm-{i386,x86_64}/early_printk.h
- run the setup in start_kernel() before setup_arch()

arch/i386/Kconfig                 |   20 +++
arch/x86_64/kernel/early_printk.c |  218 -----------------------------
arch/x86_64/kernel/head64.c       |    5
include/asm-i386/early_printk.h   |    8 +
include/asm-x86_64/early_printk.h |    8 +
include/linux/early_printk.h      |   47 ++++++++
init/main.c                       |    1
kernel/Makefile                   |    1
kernel/early_printk.c             |  209 +++++++++++++++++++++++++++++
 9 files changed, 295 insertions(+), 222 deletions(-)

Dave Hansen
[email protected]

diff -urN linux-2.5.47-clean/arch/i386/Kconfig linux-2.5.47/arch/i386/Kconfig
@@ -1588,6 +1588,26 @@
 	  keys are documented in <file:Documentation/sysrq.txt>. Don't say Y
 	  unless you really know what this hack does.
 
+config EARLY_PRINTK
+	bool "Early console support"
+	default n
+	depends on DEBUG_KERNEL
+	help
+	  Write kernel log output directly into the VGA buffer or serial port. 
+	  This is useful for kernel debugging when your machine crashes very 
+	  early before the console code is initialized. For normal operation 
+	  it is not recommended because it looks ugly and doesn't cooperate 
+	  with klogd/syslogd or the X server.You should normally N here, 
+	  unless you want to debug such a crash.
+
+	  Syntax: earlyprintk=vga
+		  earlyprintk=serial[,ttySn[,baudrate]] 
+	  Append ,keep to not disable it when the real console takes over.
+	  Only vga or serial at a time, not both.
+	  Currently only ttyS0 and ttyS1 are supported. 
+	  Interaction with the standard serial driver is not very good. 
+	  The VGA output is eventually overwritten by the real console.
+
 config DEBUG_SPINLOCK
 	bool "Spinlock debugging"
 	depends on DEBUG_KERNEL

diff -urN linux-2.5.47-clean/arch/x86_64/kernel/head64.c linux-2.5.47/arch/x86_64/kernel/head64.c
@@ -70,7 +70,7 @@
 	boot_cpu_data.x86_mask = eax & 0xf;
 }
 
-extern void start_kernel(void), pda_init(int), setup_early_printk(char *); 
+extern void start_kernel(void), pda_init(int); 
 extern int disable_apic;
 
 void __init x86_64_start_kernel(char * real_mode_data)

@@ -80,9 +80,6 @@
 	clear_bss();
 	pda_init(0);
 	copy_bootdata(real_mode_data);
-	s = strstr(saved_command_line, "earlyprintk="); 
-	if (s != NULL)
-		setup_early_printk(s+12); 
 #ifdef CONFIG_X86_IO_APIC
 	if (strstr(saved_command_line, "disableapic"))
 		disable_apic = 1;

diff -urN linux-2.5.47-clean/include/asm-i386/early_printk.h linux-2.5.47/include/asm-i386/early_printk.h
@@ -0,0 +1,8 @@
+#ifndef __EARLY_PRINTK_H_I386_
+#define __EARLY_PRINTK_H_i386_
+
+#define VGABASE  0xB8000
+#define SERIAL_BASES { 0x3f8, 0x2f8 }
+#define SERIAL_BASES_LEN 2
+
+#endif

diff -urN linux-2.5.47-clean/include/asm-x86_64/early_printk.h linux-2.5.47/include/asm-x86_64/early_printk.h
@@ -0,0 +1,8 @@
+#ifdef __EARLY_PRINTK_H_X86_64_
+#define __EARLY_PRINTK_H_X86_64_
+
+#define VGABASE	0xffffffff800b8000UL
+#define SERIAL_BASES { 0x3f8, 0x2f8 }
+#define SERIAL_BASES_LEN 2
+
+#endif

Binary files linux-2.5.47-clean/include/linux/.early_printk.h.swp and linux-2.5.47/include/linux/.early_printk.h.swp differ

diff -urN linux-2.5.47-clean/init/main.c linux-2.5.47/init/main.c
@@ -33,6 +33,7 @@
 #include <linux/workqueue.h>
 #include <linux/profile.h>
 #include <linux/rcupdate.h>
+#include <linux/early_printk.h>
 
 #include <asm/io.h>
 #include <asm/bugs.h>

@@ -387,6 +388,7 @@
  */
 	lock_kernel();
 	printk(linux_banner);
+	setup_early_printk(&command_line);
 	setup_arch(&command_line);
 	setup_per_cpu_areas();
 	build_all_zonelists();

diff -urN linux-2.5.47-clean/kernel/Makefile linux-2.5.47/kernel/Makefile
@@ -21,6 +21,7 @@
 obj-$(CONFIG_CPU_FREQ) += cpufreq.o
 obj-$(CONFIG_BSD_PROCESS_ACCT) += acct.o
 obj-$(CONFIG_SOFTWARE_SUSPEND) += suspend.o
+obj-$(CONFIG_EARLY_PRINTK) += early_printk.o
 
 ifneq ($(CONFIG_IA64),y)
 # According to Alan Modra <[email protected]>, the -fno-omit-frame-pointer is

Looks pretty decent to me... minor comments:

* might be a good idea to match defaults with the defaults of the normal serial console. IIRC that's 38400,N,8,1. which brings up a tangent, 57600 or 115200 is probably preferred these days, but that requires a later patch to update serial console and early serial console to be faster

* VGA base might want to be a variable, though I have not convinced myself that such is useful

That's overkill. Most architectures have an early_printk equivalent in firmware. Only i386 and x86-64 are not lucky enough to have one that is usable from the CPU mode linux uses.

-Andi

Perhaps s/early_printk/prom_printf/ would pick up a few arches for free.

Bill

Followup to:  <[email protected]>
By author:    Andi Kleen <[email protected]>
In newsgroup: linux.dev.kernel

That's a pretty big assumption. I wouldn't hold that to be necessarily true.

-hpa

<[email protected]> at work, <[email protected]> in private!
"Unix gives you enough rope to shoot yourself in the foot."
http://www.zytor.com/~hpa/puzzle.txt	<[email protected]>

Why is it necessary to reimplement serial console for each arch rather than just registering the generic serial console as early as possible from arch-specific code?

There's no _reason_ to wait for console_init(), for most consoles.

--
dwmw2

The generic serial driver needs lots of infrastructure that is not present in early boot. Also the early_printk serial driver is polled, not interrupt driven. The whole point of early_printk is that you can debug from line one of your code, so it is essential that it is self contained as far as possible.

-Andi

No, the generic serial _console_ is polled and should require no more infrastructure than the early_console version which it strongly resembles.

After all, it's generally called from console_init(), at which point there's little more infrastructure available than in setup_arch() _anyway_.

You should happily be able to call serial8250_console_init() earlier then console_init() currently does.

(Btw, if you're playing with this code I posted a cleanup to all the horrible #ifdefs in console_init, converting it to the __initcall mechanism.
http://marc.theaimsgroup.com/?l=linux-kernel&m=103660078206270&w=2 )

--
dwmw2