From: Andi Kleen <ak@muc.de>
Add a new "M" tainted bit in oopses for machine checks.
This shows in a oops when the machine had already a machine check. That
information may be useful to see if the machine has hardware trouble and if
it's really worth to invest much time into a mysterious oops.
Support for i386 and x86-64 so far.
Signed-off-by: Andrew Morton <akpm@osdl.org>
---
25-akpm/arch/i386/kernel/cpu/mcheck/k7.c | 2 ++
25-akpm/arch/i386/kernel/cpu/mcheck/p4.c | 2 ++
25-akpm/arch/i386/kernel/cpu/mcheck/p5.c | 2 ++
25-akpm/arch/i386/kernel/cpu/mcheck/p6.c | 2 ++
25-akpm/arch/i386/kernel/cpu/mcheck/winchip.c | 1 +
25-akpm/arch/x86_64/kernel/mce.c | 1 +
25-akpm/include/linux/kernel.h | 1 +
25-akpm/kernel/panic.c | 4 +++-
8 files changed, 14 insertions(+), 1 deletion(-)
diff -puN arch/i386/kernel/cpu/mcheck/k7.c~add-tainted-bit-for-machine-checks arch/i386/kernel/cpu/mcheck/k7.c
--- 25/arch/i386/kernel/cpu/mcheck/k7.c~add-tainted-bit-for-machine-checks 2004-09-30 22:37:11.400017152 -0700
+++ 25-akpm/arch/i386/kernel/cpu/mcheck/k7.c 2004-09-30 22:37:11.414015024 -0700
@@ -64,6 +64,8 @@ static asmlinkage void k7_machine_check(
printk (KERN_EMERG "Attempting to continue.\n");
mcgstl &= ~(1<<2);
wrmsr (MSR_IA32_MCG_STATUS,mcgstl, mcgsth);
+
+ tainted |= TAINT_MACHINE_CHECK;
}
diff -puN arch/i386/kernel/cpu/mcheck/p4.c~add-tainted-bit-for-machine-checks arch/i386/kernel/cpu/mcheck/p4.c
--- 25/arch/i386/kernel/cpu/mcheck/p4.c~add-tainted-bit-for-machine-checks 2004-09-30 22:37:11.402016848 -0700
+++ 25-akpm/arch/i386/kernel/cpu/mcheck/p4.c 2004-09-30 22:37:11.414015024 -0700
@@ -226,6 +226,8 @@ static asmlinkage void intel_machine_che
}
mcgstl &= ~(1<<2);
wrmsr (MSR_IA32_MCG_STATUS,mcgstl, mcgsth);
+
+ tainted |= TAINT_MACHINE_CHECK;
}
diff -puN arch/i386/kernel/cpu/mcheck/p5.c~add-tainted-bit-for-machine-checks arch/i386/kernel/cpu/mcheck/p5.c
--- 25/arch/i386/kernel/cpu/mcheck/p5.c~add-tainted-bit-for-machine-checks 2004-09-30 22:37:11.403016696 -0700
+++ 25-akpm/arch/i386/kernel/cpu/mcheck/p5.c 2004-09-30 22:37:11.415014872 -0700
@@ -25,6 +25,8 @@ static asmlinkage void pentium_machine_c
printk(KERN_EMERG "CPU#%d: Machine Check Exception: 0x%8X (type 0x%8X).\n", smp_processor_id(), loaddr, lotype);
if(lotype&(1<<5))
printk(KERN_EMERG "CPU#%d: Possible thermal failure (CPU on fire ?).\n", smp_processor_id());
+
+ tainted |= TAINT_MACHINE_CHECK;
}
/* Set up machine check reporting for processors with Intel style MCE */
diff -puN arch/i386/kernel/cpu/mcheck/p6.c~add-tainted-bit-for-machine-checks arch/i386/kernel/cpu/mcheck/p6.c
--- 25/arch/i386/kernel/cpu/mcheck/p6.c~add-tainted-bit-for-machine-checks 2004-09-30 22:37:11.405016392 -0700
+++ 25-akpm/arch/i386/kernel/cpu/mcheck/p6.c 2004-09-30 22:37:11.415014872 -0700
@@ -76,6 +76,8 @@ static asmlinkage void intel_machine_che
}
mcgstl &= ~(1<<2);
wrmsr (MSR_IA32_MCG_STATUS,mcgstl, mcgsth);
+
+ tainted |= TAINT_MACHINE_CHECK;
}
/* Set up machine check reporting for processors with Intel style MCE */
diff -puN arch/i386/kernel/cpu/mcheck/winchip.c~add-tainted-bit-for-machine-checks arch/i386/kernel/cpu/mcheck/winchip.c
--- 25/arch/i386/kernel/cpu/mcheck/winchip.c~add-tainted-bit-for-machine-checks 2004-09-30 22:37:11.406016240 -0700
+++ 25-akpm/arch/i386/kernel/cpu/mcheck/winchip.c 2004-09-30 22:37:11.415014872 -0700
@@ -19,6 +19,7 @@
static asmlinkage void winchip_machine_check(struct pt_regs * regs, long error_code)
{
printk(KERN_EMERG "CPU0: Machine Check Exception.\n");
+ tainted |= TAINT_MACHINE_CHECK;
}
/* Set up machine check reporting on the Winchip C6 series */
diff -puN arch/x86_64/kernel/mce.c~add-tainted-bit-for-machine-checks arch/x86_64/kernel/mce.c
--- 25/arch/x86_64/kernel/mce.c~add-tainted-bit-for-machine-checks 2004-09-30 22:37:11.407016088 -0700
+++ 25-akpm/arch/x86_64/kernel/mce.c 2004-09-30 22:37:11.416014720 -0700
@@ -227,6 +227,7 @@ void do_machine_check(struct pt_regs * r
}
out:
+ tainted |= TAINT_MACHINE_CHECK;
/* Last thing done in the machine check exception to clear state. */
wrmsrl(MSR_IA32_MCG_STATUS, 0);
}
diff -puN include/linux/kernel.h~add-tainted-bit-for-machine-checks include/linux/kernel.h
--- 25/include/linux/kernel.h~add-tainted-bit-for-machine-checks 2004-09-30 22:37:11.409015784 -0700
+++ 25-akpm/include/linux/kernel.h 2004-09-30 22:37:11.417014568 -0700
@@ -146,6 +146,7 @@ extern enum system_states {
SYSTEM_RESTART,
} system_state;
+#define TAINT_MACHINE_CHECK (1<<10)
#define TAINT_PROPRIETARY_MODULE (1<<0)
#define TAINT_FORCED_MODULE (1<<1)
#define TAINT_UNSAFE_SMP (1<<2)
diff -puN kernel/panic.c~add-tainted-bit-for-machine-checks kernel/panic.c
--- 25/kernel/panic.c~add-tainted-bit-for-machine-checks 2004-09-30 22:37:11.410015632 -0700
+++ 25-akpm/kernel/panic.c 2004-09-30 22:37:11.417014568 -0700
@@ -117,6 +117,7 @@ EXPORT_SYMBOL(panic);
* 'P' - Proprietary module has been loaded.
* 'F' - Module has been forcibly loaded.
* 'S' - SMP with CPUs not designed for SMP.
+ * 'M' - Machine had a machine check experience.
*
* The string is overwritten by the next call to print_taint().
*/
@@ -125,7 +126,8 @@ const char *print_tainted(void)
{
static char buf[20];
if (tainted) {
- snprintf(buf, sizeof(buf), "Tainted: %c%c%c",
+ snprintf(buf, sizeof(buf), "Tainted: %c%c%c%c",
+ tainted & TAINT_MACHINE_CHECK ? 'M' : ' ',
tainted & TAINT_PROPRIETARY_MODULE ? 'P' : 'G',
tainted & TAINT_FORCED_MODULE ? 'F' : ' ',
tainted & TAINT_UNSAFE_SMP ? 'S' : ' ');
_