[patch 2.6.16-rc6-mm2] x86: add NUMA to oops messages

[patch 2.6.16-rc6-mm2] x86: add NUMA to oops messages

[Chuck Ebbert]

Add "NUMA" to x86 oops printouts to help with debugging.  Use vermagic.h
defines to clean up the code, suggested by Arjan van de Ven.

Signed-off-by: Chuck Ebbert <76306.1226@compuserve.com>

 arch/i386/kernel/traps.c   |   15 ++++-----------
 arch/x86_64/kernel/traps.c |   13 +++----------
 include/linux/vermagic.h   |   15 ++++++++++++++-
 3 files changed, 21 insertions(+), 22 deletions(-)

--- 2.6.17-rc6-mm2-32.orig/arch/i386/kernel/traps.c
+++ 2.6.17-rc6-mm2-32/arch/i386/kernel/traps.c
@@ -29,6 +29,7 @@
 #include <linux/kprobes.h>
 #include <linux/kexec.h>
 #include <linux/unwind.h>
+#include <linux/vermagic.h>
 
 #ifdef CONFIG_EISA
 #include <linux/ioport.h>
@@ -391,17 +392,9 @@ void die(const char * str, struct pt_reg
 		handle_BUG(regs);
 		printk(KERN_EMERG "%s: %04lx [#%d]\n", str, err & 0xffff,
 				++die_counter);
-		printk(KERN_EMERG "%dK_STACKS ", THREAD_SIZE / 1024);
-#ifdef CONFIG_PREEMPT
-		printk("PREEMPT ");
-#endif
-#ifdef CONFIG_SMP
-		printk("SMP ");
-#endif
-#ifdef CONFIG_DEBUG_PAGEALLOC
-		printk("DEBUG_PAGEALLOC");
-#endif
-		printk("\n");
+		printk(KERN_EMERG MODULE_VERMAGIC_PREEMPT MODULE_VERMAGIC_SMP
+		       VERMAGIC_NUMA VERMAGIC_DEBUG_PAGEALLOC);
+		printk("%dK_STACKS\n", THREAD_SIZE / 1024);
 		sysfs_printk_last_file();
 		if (notify_die(DIE_OOPS, str, regs, err,
 			    current->thread.trap_no, SIGSEGV) != NOTIFY_STOP) {
--- 2.6.17-rc6-mm2-32.orig/arch/x86_64/kernel/traps.c
+++ 2.6.17-rc6-mm2-32/arch/x86_64/kernel/traps.c
@@ -30,6 +30,7 @@
 #include <linux/kprobes.h>
 #include <linux/kexec.h>
 #include <linux/unwind.h>
+#include <linux/vermagic.h>
 
 #include <asm/system.h>
 #include <asm/uaccess.h>
@@ -528,16 +529,8 @@ void __kprobes __die(const char * str, s
 {
 	static int die_counter;
 	printk(KERN_EMERG "%s: %04lx [%u] ", str, err & 0xffff,++die_counter);
-#ifdef CONFIG_PREEMPT
-	printk("PREEMPT ");
-#endif
-#ifdef CONFIG_SMP
-	printk("SMP ");
-#endif
-#ifdef CONFIG_DEBUG_PAGEALLOC
-	printk("DEBUG_PAGEALLOC");
-#endif
-	printk("\n");
+	printk(MODULE_VERMAGIC_PREEMPT MODULE_VERMAGIC_SMP
+	       VERMAGIC_NUMA VERMAGIC_DEBUG_PAGEALLOC "\n");
 	sysfs_printk_last_file();
 	notify_die(DIE_OOPS, str, regs, err, current->thread.trap_no, SIGSEGV);
 	show_registers(regs);
--- 2.6.17-rc6-mm2-32.orig/include/linux/vermagic.h
+++ 2.6.17-rc6-mm2-32/include/linux/vermagic.h
@@ -8,7 +8,7 @@
 #define MODULE_VERMAGIC_SMP ""
 #endif
 #ifdef CONFIG_PREEMPT
-#define MODULE_VERMAGIC_PREEMPT "preempt "
+#define MODULE_VERMAGIC_PREEMPT "PREEMPT "
 #else
 #define MODULE_VERMAGIC_PREEMPT ""
 #endif
@@ -26,3 +26,16 @@
 	MODULE_VERMAGIC_SMP MODULE_VERMAGIC_PREEMPT 			\
 	MODULE_VERMAGIC_MODULE_UNLOAD MODULE_ARCH_VERMAGIC 		\
 	"gcc-" __stringify(__GNUC__) "." __stringify(__GNUC_MINOR__)
+
+/* for printing in debug routines */
+#ifdef CONFIG_DEBUG_PAGEALLOC
+#define VERMAGIC_DEBUG_PAGEALLOC "DEBUG_PAGEALLOC "
+#else
+#define VERMAGIC_DEBUG_PAGEALLOC ""
+#endif
+
+#ifdef CONFIG_NUMA
+#define VERMAGIC_NUMA "NUMA "
+#else
+#define VERMAGIC_NUMA ""
+#endif
-- 
Chuck
 "You can't read a newspaper if you can't read."  --George W. Bush

Re: [patch 2.6.16-rc6-mm2] x86: add NUMA to oops messages

[Andi Kleen]

On Sunday 18 June 2006 10:13, Chuck Ebbert wrote:
> Add "NUMA" to x86 oops printouts to help with debugging.  Use vermagic.h
> defines to clean up the code, suggested by Arjan van de Ven.

I don't see any particular value in printing NUMA here,
and putting half of .config into the oops doesn't seem
like a good long term strategy.

-Andi

  

Re: [patch 2.6.16-rc6-mm2] x86: add NUMA to oops messages

[Chuck Ebbert]

In-Reply-To: <200606181059.44576.ak@suse.de>

On Sun, 18 Jun 2006 10:59:44 +0200, Andi Kleen wrote:

> On Sunday 18 June 2006 10:13, Chuck Ebbert wrote:
> > Add "NUMA" to x86 oops printouts to help with debugging.  Use vermagic.h
> > defines to clean up the code, suggested by Arjan van de Ven.
> 
> I don't see any particular value in printing NUMA here,

It helps a lot when trying to decode oopses posted to the list, as the
NUMA kernel is very different now that the slab allocator and scheduler
are NUMA-aware.

 $ fgrep 'def CONFIG_NUMA' mm/*.c kernel/*.c include/linux/* | wc -l
 58
 $ fgrep 'def CONFIG_SMP' mm/*.c kernel/*.c include/linux/* | wc -l
 81

> and putting half of .config into the oops doesn't seem
> like a good long term strategy.

It's just this one thing; I can't think of anything else to add...

-- 
Chuck
 "You can't read a newspaper if you can't read."  --George W. Bush

Re: [patch 2.6.16-rc6-mm2] x86: add NUMA to oops messages

[Arjan van de Ven]

Chuck Ebbert wrote:
>> and putting half of .config into the oops doesn't seem
>> like a good long term strategy.
> 
> It's just this one thing; I can't think of anything else to add...
> 

if it's compact I'd actually put the entire vermagic line there...
it'll show compiler, preempt and a whole lot of other "distinctive" features

Re: [patch 2.6.16-rc6-mm2] x86: add NUMA to oops messages

[Randy.Dunlap]

On Sun, 18 Jun 2006 14:48:27 +0200 Arjan van de Ven wrote:

> Chuck Ebbert wrote:
> >> and putting half of .config into the oops doesn't seem
> >> like a good long term strategy.
> > 
> > It's just this one thing; I can't think of anything else to add...
> > 
> 
> if it's compact I'd actually put the entire vermagic line there...
> it'll show compiler, preempt and a whole lot of other "distinctive" features

Yep, I like that idea too.

---
~Randy

Re: [patch 2.6.16-rc6-mm2] x86: add NUMA to oops messages

[Andreas Kleen]

Am So 18.06.2006 14:13 schrieb Chuck Ebbert <76306.1226@compuserve.com>:

> In-Reply-To: <200606181059.44576.ak@suse.de>
>
> On Sun, 18 Jun 2006 10:59:44 +0200, Andi Kleen wrote:
>
> > On Sunday 18 June 2006 10:13, Chuck Ebbert wrote:
> > > Add "NUMA" to x86 oops printouts to help with debugging. Use
> > > vermagic.h
> > > defines to clean up the code, suggested by Arjan van de Ven.
> >
> > I don't see any particular value in printing NUMA here,
>
> It helps a lot when trying to decode oopses posted to the list, as the
> NUMA kernel is very different now that the slab allocator and
> scheduler
> are NUMA-aware.

And there are a thousand other things that change with other
CONFIG_* options. I also can't remember a bug where
this wasn't clear.

In general the rule of thumb is: i386: NUMA is not
used and x86-64 SMP: NUMA is used.

-Andi