From f784946ded3e734524d1f48a6a8286b8a152c3b9 Mon Sep 17 00:00:00 2001 From: Hiroshi Shimamoto Date: Fri, 2 May 2008 16:45:08 -0700 Subject: x86: use per_cpu data in nmi_32 use per_cpu for per CPU data. Signed-off-by: Hiroshi Shimamoto Signed-off-by: Thomas Gleixner Signed-off-by: Ingo Molnar --- arch/x86/kernel/nmi_32.c | 29 +++++++++++++++++------------ 1 file changed, 17 insertions(+), 12 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c index 11b14bbaa61e..69bdae555c18 100644 --- a/arch/x86/kernel/nmi_32.c +++ b/arch/x86/kernel/nmi_32.c @@ -293,9 +293,9 @@ void stop_apic_nmi_watchdog(void *unused) * here too!] */ -static unsigned int - last_irq_sums [NR_CPUS], - alert_counter [NR_CPUS]; +static DEFINE_PER_CPU(unsigned, last_irq_sum); +static DEFINE_PER_CPU(local_t, alert_counter); +static DEFINE_PER_CPU(int, nmi_touch); void touch_nmi_watchdog(void) { @@ -303,12 +303,13 @@ void touch_nmi_watchdog(void) unsigned cpu; /* - * Just reset the alert counters, (other CPUs might be - * spinning on locks we hold): + * Tell other CPUs to reset their alert counters. We cannot + * do it ourselves because the alert count increase is not + * atomic. */ for_each_present_cpu(cpu) { - if (alert_counter[cpu]) - alert_counter[cpu] = 0; + if (per_cpu(nmi_touch, cpu) != 1) + per_cpu(nmi_touch, cpu) = 1; } } @@ -358,22 +359,26 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) */ sum = per_cpu(irq_stat, cpu).apic_timer_irqs + per_cpu(irq_stat, cpu).irq0_irqs; + if (__get_cpu_var(nmi_touch)) { + __get_cpu_var(nmi_touch) = 0; + touched = 1; + } /* if the none of the timers isn't firing, this cpu isn't doing much */ - if (!touched && last_irq_sums[cpu] == sum) { + if (!touched && __get_cpu_var(last_irq_sum) == sum) { /* * Ayiee, looks like this CPU is stuck ... * wait a few IRQs (5 seconds) before doing the oops ... */ - alert_counter[cpu]++; - if (alert_counter[cpu] == 5*nmi_hz) + local_inc(&__get_cpu_var(alert_counter)); + if (local_read(&__get_cpu_var(alert_counter)) == 5*nmi_hz) /* * die_nmi will return ONLY if NOTIFY_STOP happens.. */ die_nmi(regs, "BUG: NMI Watchdog detected LOCKUP"); } else { - last_irq_sums[cpu] = sum; - alert_counter[cpu] = 0; + __get_cpu_var(last_irq_sum) = sum; + local_set(&__get_cpu_var(alert_counter), 0); } /* see if the nmi watchdog went off */ if (!__get_cpu_var(wd_enabled)) -- cgit v1.2.3 From f59a9310b97879159ef4d25227bc270278f1ee71 Mon Sep 17 00:00:00 2001 From: Hiroshi Shimamoto Date: Fri, 2 May 2008 16:44:52 -0700 Subject: x86: nmi and irq counters are unsigned in nmi_64.c __nmi_count, apic_timer_irqs and irq0_irqs are unsigned. Signed-off-by: Hiroshi Shimamoto Signed-off-by: Thomas Gleixner Signed-off-by: Ingo Molnar --- arch/x86/kernel/nmi_64.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_64.c b/arch/x86/kernel/nmi_64.c index 5a29ded994fa..33cb8ecbb6dd 100644 --- a/arch/x86/kernel/nmi_64.c +++ b/arch/x86/kernel/nmi_64.c @@ -79,7 +79,7 @@ static __init void nmi_cpu_busy(void *data) int __init check_nmi_watchdog(void) { - int *prev_nmi_count; + unsigned int *prev_nmi_count; int cpu; if ((nmi_watchdog == NMI_NONE) || (nmi_watchdog == NMI_DISABLED)) @@ -316,7 +316,7 @@ EXPORT_SYMBOL(touch_nmi_watchdog); notrace __kprobes int nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) { - int sum; + unsigned int sum; int touched = 0; int cpu = smp_processor_id(); int rc = 0; -- cgit v1.2.3 From ddca03c98a7f7ad5ab09967ff52d4ed60358c896 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:31 +0400 Subject: x86: nmi - unify die_nmi() interface By slightly changing 32bit mode die_nmi() we may unify the interface and make it common for both (32/64bit) modes Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_32.c | 7 +++---- arch/x86/kernel/traps_32.c | 8 +++++--- include/asm-x86/nmi.h | 2 +- 3 files changed, 9 insertions(+), 8 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c index 69bdae555c18..bd04a28f7a5c 100644 --- a/arch/x86/kernel/nmi_32.c +++ b/arch/x86/kernel/nmi_32.c @@ -320,8 +320,6 @@ void touch_nmi_watchdog(void) } EXPORT_SYMBOL(touch_nmi_watchdog); -extern void die_nmi(struct pt_regs *, const char *msg); - notrace __kprobes int nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) { @@ -375,7 +373,8 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) /* * die_nmi will return ONLY if NOTIFY_STOP happens.. */ - die_nmi(regs, "BUG: NMI Watchdog detected LOCKUP"); + die_nmi("BUG: NMI Watchdog detected LOCKUP", + regs, 0); } else { __get_cpu_var(last_irq_sum) = sum; local_set(&__get_cpu_var(alert_counter), 0); @@ -406,7 +405,7 @@ static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu) char buf[64]; sprintf(buf, "NMI received for unknown reason %02x\n", reason); - die_nmi(regs, buf); + die_nmi(buf, regs, 0); return 0; } diff --git a/arch/x86/kernel/traps_32.c b/arch/x86/kernel/traps_32.c index bde6f63e15d5..f31e6651fa6f 100644 --- a/arch/x86/kernel/traps_32.c +++ b/arch/x86/kernel/traps_32.c @@ -755,9 +755,9 @@ unknown_nmi_error(unsigned char reason, struct pt_regs *regs) static DEFINE_SPINLOCK(nmi_print_lock); -void notrace __kprobes die_nmi(struct pt_regs *regs, const char *msg) +void notrace __kprobes die_nmi(char *str, struct pt_regs *regs, int do_panic) { - if (notify_die(DIE_NMIWATCHDOG, msg, regs, 0, 2, SIGINT) == NOTIFY_STOP) + if (notify_die(DIE_NMIWATCHDOG, str, regs, 0, 2, SIGINT) == NOTIFY_STOP) return; spin_lock(&nmi_print_lock); @@ -766,10 +766,12 @@ void notrace __kprobes die_nmi(struct pt_regs *regs, const char *msg) * to get a message out: */ bust_spinlocks(1); - printk(KERN_EMERG "%s", msg); + printk(KERN_EMERG "%s", str); printk(" on CPU%d, ip %08lx, registers:\n", smp_processor_id(), regs->ip); show_registers(regs); + if (do_panic) + panic("Non maskable interrupt"); console_silent(); spin_unlock(&nmi_print_lock); bust_spinlocks(0); diff --git a/include/asm-x86/nmi.h b/include/asm-x86/nmi.h index 8455bf36d8df..7cd5b6a88138 100644 --- a/include/asm-x86/nmi.h +++ b/include/asm-x86/nmi.h @@ -38,12 +38,12 @@ static inline void unset_nmi_pm_callback(struct pm_dev *dev) #ifdef CONFIG_X86_64 extern void default_do_nmi(struct pt_regs *); -extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); extern void nmi_watchdog_default(void); #else #define nmi_watchdog_default() do {} while (0) #endif +extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); extern int check_nmi_watchdog(void); extern int nmi_watchdog_enabled; extern int avail_to_resrv_perfctr_nmi_bit(unsigned int); -- cgit v1.2.3 From e56b3a12c45e439169f2d1f7194be397667320a6 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:32 +0400 Subject: x86: nmi - die_nmi() output message unification Make 64bit die_nmi() to produce the same message as 32bit mode has Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_64.c | 4 ++-- arch/x86/kernel/traps_64.c | 4 +++- 2 files changed, 5 insertions(+), 3 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_64.c b/arch/x86/kernel/nmi_64.c index 33cb8ecbb6dd..1a98705270a2 100644 --- a/arch/x86/kernel/nmi_64.c +++ b/arch/x86/kernel/nmi_64.c @@ -358,8 +358,8 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) */ local_inc(&__get_cpu_var(alert_counter)); if (local_read(&__get_cpu_var(alert_counter)) == 5*nmi_hz) - die_nmi("NMI Watchdog detected LOCKUP on CPU %d\n", regs, - panic_on_timeout); + die_nmi("NMI Watchdog detected LOCKUP", + regs, panic_on_timeout); } else { __get_cpu_var(last_irq_sum) = sum; local_set(&__get_cpu_var(alert_counter), 0); diff --git a/arch/x86/kernel/traps_64.c b/arch/x86/kernel/traps_64.c index adff76ea97c4..6a048ce3c787 100644 --- a/arch/x86/kernel/traps_64.c +++ b/arch/x86/kernel/traps_64.c @@ -614,7 +614,9 @@ die_nmi(char *str, struct pt_regs *regs, int do_panic) * We are in trouble anyway, lets at least try * to get a message out. */ - printk(str, smp_processor_id()); + printk(KERN_EMERG "%s", str); + printk(" on CPU%d, ip %08lx, registers:\n", + smp_processor_id(), regs->ip); show_registers(regs); if (kexec_should_crash(current)) crash_kexec(regs); -- cgit v1.2.3 From c6425b9f143a75bbcd0a7684b4df40e20d0b2f32 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:33 +0400 Subject: x86: move do_nmi(), stop_nmi() and restart_nmi() to traps_64.c traps_32.c already holds these functions so do the same for traps_64.c Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_64.c | 25 ------------------------- arch/x86/kernel/traps_64.c | 24 ++++++++++++++++++++++++ 2 files changed, 24 insertions(+), 25 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_64.c b/arch/x86/kernel/nmi_64.c index 1a98705270a2..c1ea671525e5 100644 --- a/arch/x86/kernel/nmi_64.c +++ b/arch/x86/kernel/nmi_64.c @@ -30,7 +30,6 @@ int unknown_nmi_panic; int nmi_watchdog_enabled; -int panic_on_unrecovered_nmi; static cpumask_t backtrace_mask = CPU_MASK_NONE; @@ -383,30 +382,6 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) return rc; } -static unsigned ignore_nmis; - -asmlinkage notrace __kprobes void -do_nmi(struct pt_regs *regs, long error_code) -{ - nmi_enter(); - add_pda(__nmi_count,1); - if (!ignore_nmis) - default_do_nmi(regs); - nmi_exit(); -} - -void stop_nmi(void) -{ - acpi_nmi_disable(); - ignore_nmis++; -} - -void restart_nmi(void) -{ - ignore_nmis--; - acpi_nmi_enable(); -} - #ifdef CONFIG_SYSCTL static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu) diff --git a/arch/x86/kernel/traps_64.c b/arch/x86/kernel/traps_64.c index 6a048ce3c787..e4a380797dc1 100644 --- a/arch/x86/kernel/traps_64.c +++ b/arch/x86/kernel/traps_64.c @@ -76,7 +76,9 @@ asmlinkage void alignment_check(void); asmlinkage void machine_check(void); asmlinkage void spurious_interrupt_bug(void); +int panic_on_unrecovered_nmi; static unsigned int code_bytes = 64; +static unsigned ignore_nmis; static inline void conditional_sti(struct pt_regs *regs) { @@ -867,6 +869,28 @@ asmlinkage notrace __kprobes void default_do_nmi(struct pt_regs *regs) io_check_error(reason, regs); } +asmlinkage notrace __kprobes void +do_nmi(struct pt_regs *regs, long error_code) +{ + nmi_enter(); + add_pda(__nmi_count, 1); + if (!ignore_nmis) + default_do_nmi(regs); + nmi_exit(); +} + +void stop_nmi(void) +{ + acpi_nmi_disable(); + ignore_nmis++; +} + +void restart_nmi(void) +{ + ignore_nmis--; + acpi_nmi_enable(); +} + /* runs on IST stack. */ asmlinkage void __kprobes do_int3(struct pt_regs * regs, long error_code) { -- cgit v1.2.3 From d1b946b97d71423f365fa797d1428e1847c0bec1 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:34 +0400 Subject: x86: nmi_32.c - add "panic" option Allow to pass "panic" option in 32bit mode Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_32.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c index bd04a28f7a5c..4437fe1edabc 100644 --- a/arch/x86/kernel/nmi_32.c +++ b/arch/x86/kernel/nmi_32.c @@ -42,6 +42,7 @@ static cpumask_t backtrace_mask = CPU_MASK_NONE; * 0: the lapic NMI watchdog is disabled, but can be enabled */ atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ +static int panic_on_timeout; unsigned int nmi_watchdog = NMI_DEFAULT; static unsigned int nmi_hz = HZ; @@ -140,6 +141,14 @@ static int __init setup_nmi_watchdog(char *str) { int nmi; + if (!strncmp(str, "panic", 5)) { + panic_on_timeout = 1; + str = strchr(str, ','); + if (!str) + return 1; + ++str; + } + get_option(&str, &nmi); if ((nmi >= NMI_INVALID) || (nmi < NMI_NONE)) @@ -374,7 +383,7 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) * die_nmi will return ONLY if NOTIFY_STOP happens.. */ die_nmi("BUG: NMI Watchdog detected LOCKUP", - regs, 0); + regs, panic_on_timeout); } else { __get_cpu_var(last_irq_sum) = sum; local_set(&__get_cpu_var(alert_counter), 0); -- cgit v1.2.3 From 4b82b277707a39b97271439c475f186f63ec4692 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:35 +0400 Subject: x86: nmi_32.c - add nmi_watchdog_default helper Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_32.c | 19 +++++++++++++------ include/asm-x86/nmi.h | 4 +--- 2 files changed, 14 insertions(+), 9 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c index 4437fe1edabc..7714e8478213 100644 --- a/arch/x86/kernel/nmi_32.c +++ b/arch/x86/kernel/nmi_32.c @@ -51,6 +51,17 @@ static DEFINE_PER_CPU(short, wd_enabled); static int endflag __initdata = 0; +/* Run after command line and cpu_init init, but before all other checks */ +void nmi_watchdog_default(void) +{ + if (nmi_watchdog != NMI_DEFAULT) + return; + if (lapic_watchdog_ok()) + nmi_watchdog = NMI_LOCAL_APIC; + else + nmi_watchdog = NMI_IO_APIC; +} + #ifdef CONFIG_SMP /* The performance counters used by NMI_LOCAL_APIC don't trigger when * the CPU is idle. To make sure the NMI watchdog really ticks on all @@ -437,12 +448,8 @@ int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file, return -EIO; } - if (nmi_watchdog == NMI_DEFAULT) { - if (lapic_watchdog_ok()) - nmi_watchdog = NMI_LOCAL_APIC; - else - nmi_watchdog = NMI_IO_APIC; - } + /* if nmi_watchdog is not set yet, then set it */ + nmi_watchdog_default(); if (nmi_watchdog == NMI_LOCAL_APIC) { if (nmi_watchdog_enabled) diff --git a/include/asm-x86/nmi.h b/include/asm-x86/nmi.h index 7cd5b6a88138..1e8f34d7ab68 100644 --- a/include/asm-x86/nmi.h +++ b/include/asm-x86/nmi.h @@ -38,11 +38,9 @@ static inline void unset_nmi_pm_callback(struct pm_dev *dev) #ifdef CONFIG_X86_64 extern void default_do_nmi(struct pt_regs *); -extern void nmi_watchdog_default(void); -#else -#define nmi_watchdog_default() do {} while (0) #endif +extern void nmi_watchdog_default(void); extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); extern int check_nmi_watchdog(void); extern int nmi_watchdog_enabled; -- cgit v1.2.3 From ad63ba169d45ccb6594eb35a6c31c421fe70ff45 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:36 +0400 Subject: x86: nmi_32/64.c - use apic_write_around instead of apic_write apic_write_around will be expanded to apic_write in 64bit mode anyway. Only a few CPUs (well, old CPUs to be precise) requires such an action. In general it should not hurt and could be cleaned up for apic_write (just in case) Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_32.c | 2 +- arch/x86/kernel/nmi_64.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c index 7714e8478213..ec5842b4dd68 100644 --- a/arch/x86/kernel/nmi_32.c +++ b/arch/x86/kernel/nmi_32.c @@ -248,7 +248,7 @@ void acpi_nmi_enable(void) static void __acpi_nmi_disable(void *__unused) { - apic_write(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); + apic_write_around(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); } /* diff --git a/arch/x86/kernel/nmi_64.c b/arch/x86/kernel/nmi_64.c index c1ea671525e5..f546e3164a74 100644 --- a/arch/x86/kernel/nmi_64.c +++ b/arch/x86/kernel/nmi_64.c @@ -215,7 +215,7 @@ late_initcall(init_lapic_nmi_sysfs); static void __acpi_nmi_enable(void *__unused) { - apic_write(APIC_LVT0, APIC_DM_NMI); + apic_write_around(APIC_LVT0, APIC_DM_NMI); } /* @@ -229,7 +229,7 @@ void acpi_nmi_enable(void) static void __acpi_nmi_disable(void *__unused) { - apic_write(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); + apic_write_around(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); } /* -- cgit v1.2.3 From 6c8decdf14b17d42f6eb817d310642f4d6598a19 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:37 +0400 Subject: x86: nmi_32.c - unknown_nmi_panic_callback should always panic Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_32.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c index ec5842b4dd68..d0b0684bde9e 100644 --- a/arch/x86/kernel/nmi_32.c +++ b/arch/x86/kernel/nmi_32.c @@ -425,7 +425,7 @@ static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu) char buf[64]; sprintf(buf, "NMI received for unknown reason %02x\n", reason); - die_nmi(buf, regs, 0); + die_nmi(buf, regs, 1); /* Always panic here */ return 0; } -- cgit v1.2.3 From 96f9dcb10755e96eae706b9e869c0dc25adb3d74 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:38 +0400 Subject: x86: nmi_64.c - use for_each_possible_cpu helper Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_64.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_64.c b/arch/x86/kernel/nmi_64.c index f546e3164a74..d98b21cd4928 100644 --- a/arch/x86/kernel/nmi_64.c +++ b/arch/x86/kernel/nmi_64.c @@ -98,7 +98,7 @@ int __init check_nmi_watchdog(void) smp_call_function(nmi_cpu_busy, (void *)&endflag, 0, 0); #endif - for (cpu = 0; cpu < NR_CPUS; cpu++) + for_each_possible_cpu(cpu) prev_nmi_count[cpu] = cpu_pda(cpu)->__nmi_count; local_irq_enable(); mdelay((20*1000)/nmi_hz); // wait 20 ticks -- cgit v1.2.3 From 7c2ba83f9a479eee6f302147767a30f3187fbd4b Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:39 +0400 Subject: x86: nmi_32.c cleanup - use for_each_online_cpu helper Since cpu_online_map is touched (by for_each_online_cpu) at moment when cpu_callin_map is already filled up we can get rid of its checking at all Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_32.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c index d0b0684bde9e..c6d0777aaf90 100644 --- a/arch/x86/kernel/nmi_32.c +++ b/arch/x86/kernel/nmi_32.c @@ -108,13 +108,7 @@ int __init check_nmi_watchdog(void) local_irq_enable(); mdelay((20*1000)/nmi_hz); // wait 20 ticks - for_each_possible_cpu(cpu) { -#ifdef CONFIG_SMP - /* Check cpu_callin_map here because that is set - after the timer is started. */ - if (!cpu_isset(cpu, cpu_callin_map)) - continue; -#endif + for_each_online_cpu(cpu) { if (!per_cpu(wd_enabled, cpu)) continue; if (nmi_count(cpu) - prev_nmi_count[cpu] <= 5) { -- cgit v1.2.3 From fd5cea02de100197a4c26d9e103508cf09b50a82 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:40 +0400 Subject: x86: nmi_32/64.c - add helper functions to hide arch specific data Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi_32.c | 44 ++++++++++++++++++++++++++++++++------------ arch/x86/kernel/nmi_64.c | 42 +++++++++++++++++++++++++++++++----------- 2 files changed, 63 insertions(+), 23 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c index c6d0777aaf90..a94dbedf2b74 100644 --- a/arch/x86/kernel/nmi_32.c +++ b/arch/x86/kernel/nmi_32.c @@ -51,6 +51,26 @@ static DEFINE_PER_CPU(short, wd_enabled); static int endflag __initdata = 0; +static inline unsigned int get_nmi_count(int cpu) +{ + return nmi_count(cpu); +} + +static inline int mce_in_progress(void) +{ + return 0; +} + +/* + * Take the local apic timer and PIT/HPET into account. We don't + * know which one is active, when we have highres/dyntick on + */ +static inline unsigned int get_timer_irqs(int cpu) +{ + return per_cpu(irq_stat, cpu).apic_timer_irqs + + per_cpu(irq_stat, cpu).irq0_irqs; +} + /* Run after command line and cpu_init init, but before all other checks */ void nmi_watchdog_default(void) { @@ -104,19 +124,19 @@ int __init check_nmi_watchdog(void) #endif for_each_possible_cpu(cpu) - prev_nmi_count[cpu] = nmi_count(cpu); + prev_nmi_count[cpu] = get_nmi_count(cpu); local_irq_enable(); mdelay((20*1000)/nmi_hz); // wait 20 ticks for_each_online_cpu(cpu) { if (!per_cpu(wd_enabled, cpu)) continue; - if (nmi_count(cpu) - prev_nmi_count[cpu] <= 5) { + if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5) { printk(KERN_WARNING "WARNING: CPU#%d: NMI " "appears to be stuck (%d->%d)!\n", cpu, prev_nmi_count[cpu], - nmi_count(cpu)); + get_nmi_count(cpu)); per_cpu(wd_enabled, cpu) = 0; atomic_dec(&nmi_active); } @@ -355,6 +375,13 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) touched = 1; } + sum = get_timer_irqs(cpu); + + if (__get_cpu_var(nmi_touch)) { + __get_cpu_var(nmi_touch) = 0; + touched = 1; + } + if (cpu_isset(cpu, backtrace_mask)) { static DEFINE_SPINLOCK(lock); /* Serialise the printks */ @@ -365,16 +392,9 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) cpu_clear(cpu, backtrace_mask); } - /* - * Take the local apic timer and PIT/HPET into account. We don't - * know which one is active, when we have highres/dyntick on - */ - sum = per_cpu(irq_stat, cpu).apic_timer_irqs + - per_cpu(irq_stat, cpu).irq0_irqs; - if (__get_cpu_var(nmi_touch)) { - __get_cpu_var(nmi_touch) = 0; + /* Could check oops_in_progress here too, but it's safer not to */ + if (mce_in_progress()) touched = 1; - } /* if the none of the timers isn't firing, this cpu isn't doing much */ if (!touched && __get_cpu_var(last_irq_sum) == sum) { diff --git a/arch/x86/kernel/nmi_64.c b/arch/x86/kernel/nmi_64.c index d98b21cd4928..c6802447262c 100644 --- a/arch/x86/kernel/nmi_64.c +++ b/arch/x86/kernel/nmi_64.c @@ -47,6 +47,30 @@ static unsigned int nmi_hz = HZ; static DEFINE_PER_CPU(short, wd_enabled); +static int endflag __initdata = 0; + +static inline unsigned int get_nmi_count(int cpu) +{ + return cpu_pda(cpu)->__nmi_count; +} + +static inline int mce_in_progress(void) +{ +#ifdef CONFIG_X86_MCE + return atomic_read(&mce_entry) > 0; +#endif + return 0; +} + +/* + * Take the local apic timer and PIT/HPET into account. We don't + * know which one is active, when we have highres/dyntick on + */ +static inline unsigned int get_timer_irqs(int cpu) +{ + return read_pda(apic_timer_irqs) + read_pda(irq0_irqs); +} + /* Run after command line and cpu_init init, but before all other checks */ void nmi_watchdog_default(void) { @@ -55,8 +79,6 @@ void nmi_watchdog_default(void) nmi_watchdog = NMI_NONE; } -static int endflag __initdata = 0; - #ifdef CONFIG_SMP /* The performance counters used by NMI_LOCAL_APIC don't trigger when * the CPU is idle. To make sure the NMI watchdog really ticks on all @@ -99,19 +121,19 @@ int __init check_nmi_watchdog(void) #endif for_each_possible_cpu(cpu) - prev_nmi_count[cpu] = cpu_pda(cpu)->__nmi_count; + prev_nmi_count[cpu] = get_nmi_count(cpu); local_irq_enable(); mdelay((20*1000)/nmi_hz); // wait 20 ticks for_each_online_cpu(cpu) { if (!per_cpu(wd_enabled, cpu)) continue; - if (cpu_pda(cpu)->__nmi_count - prev_nmi_count[cpu] <= 5) { + if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5) { printk(KERN_WARNING "WARNING: CPU#%d: NMI " "appears to be stuck (%d->%d)!\n", cpu, prev_nmi_count[cpu], - cpu_pda(cpu)->__nmi_count); + get_nmi_count(cpu)); per_cpu(wd_enabled, cpu) = 0; atomic_dec(&nmi_active); } @@ -327,7 +349,8 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) touched = 1; } - sum = read_pda(apic_timer_irqs) + read_pda(irq0_irqs); + sum = get_timer_irqs(cpu); + if (__get_cpu_var(nmi_touch)) { __get_cpu_var(nmi_touch) = 0; touched = 1; @@ -343,12 +366,9 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) cpu_clear(cpu, backtrace_mask); } -#ifdef CONFIG_X86_MCE - /* Could check oops_in_progress here too, but it's safer - not too */ - if (atomic_read(&mce_entry) > 0) + if (mce_in_progress()) touched = 1; -#endif + /* if the apic timer isn't firing, this cpu isn't doing much */ if (!touched && __get_cpu_var(last_irq_sum) == sum) { /* -- cgit v1.2.3 From 1798bc22b2790bf2a956588e6b17c36ef79ceff7 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 24 May 2008 19:36:41 +0400 Subject: x86: nmi_32/64.c - merge down nmi_32.c and nmi_64.c to nmi.c Signed-off-by: Cyrill Gorcunov Cc: hpa@zytor.com Cc: mingo@redhat.com Signed-off-by: Thomas Gleixner --- arch/x86/kernel/Makefile | 2 +- arch/x86/kernel/nmi.c | 533 +++++++++++++++++++++++++++++++++++++++++++++++ arch/x86/kernel/nmi_32.c | 506 -------------------------------------------- arch/x86/kernel/nmi_64.c | 477 ------------------------------------------ 4 files changed, 534 insertions(+), 984 deletions(-) create mode 100644 arch/x86/kernel/nmi.c delete mode 100644 arch/x86/kernel/nmi_32.c delete mode 100644 arch/x86/kernel/nmi_64.c (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 5e618c3b4720..d5dd6668225f 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -53,7 +53,7 @@ obj-$(CONFIG_X86_32_SMP) += smpcommon.o obj-$(CONFIG_X86_64_SMP) += tsc_sync.o smpcommon.o obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o obj-$(CONFIG_X86_MPPARSE) += mpparse.o -obj-$(CONFIG_X86_LOCAL_APIC) += apic_$(BITS).o nmi_$(BITS).o +obj-$(CONFIG_X86_LOCAL_APIC) += apic_$(BITS).o nmi.o obj-$(CONFIG_X86_IO_APIC) += io_apic_$(BITS).o obj-$(CONFIG_X86_REBOOTFIXUPS) += reboot_fixups_32.o obj-$(CONFIG_KEXEC) += machine_kexec_$(BITS).o diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c new file mode 100644 index 000000000000..69a839fc1eb0 --- /dev/null +++ b/arch/x86/kernel/nmi.c @@ -0,0 +1,533 @@ +/* + * NMI watchdog support on APIC systems + * + * Started by Ingo Molnar + * + * Fixes: + * Mikael Pettersson : AMD K7 support for local APIC NMI watchdog. + * Mikael Pettersson : Power Management for local APIC NMI watchdog. + * Mikael Pettersson : Pentium 4 support for local APIC NMI watchdog. + * Pavel Machek and + * Mikael Pettersson : PM converted to driver model. Disable/enable API. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +#include + +#include + +int unknown_nmi_panic; +int nmi_watchdog_enabled; + +static cpumask_t backtrace_mask = CPU_MASK_NONE; + +/* nmi_active: + * >0: the lapic NMI watchdog is active, but can be disabled + * <0: the lapic NMI watchdog has not been set up, and cannot + * be enabled + * 0: the lapic NMI watchdog is disabled, but can be enabled + */ +atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ +static int panic_on_timeout; + +unsigned int nmi_watchdog = NMI_DEFAULT; + +static unsigned int nmi_hz = HZ; +static DEFINE_PER_CPU(short, wd_enabled); +static int endflag __initdata = 0; + +static inline unsigned int get_nmi_count(int cpu) +{ +#ifdef CONFIG_X86_64 + return cpu_pda(cpu)->__nmi_count; +#else + return nmi_count(cpu); +#endif +} + +static inline int mce_in_progress(void) +{ +#if defined(CONFIX_X86_64) && defined(CONFIG_X86_MCE) + return atomic_read(&mce_entry) > 0; +#endif + return 0; +} + +/* + * Take the local apic timer and PIT/HPET into account. We don't + * know which one is active, when we have highres/dyntick on + */ +static inline unsigned int get_timer_irqs(int cpu) +{ +#ifdef CONFIG_X86_64 + return read_pda(apic_timer_irqs) + read_pda(irq0_irqs); +#else + return per_cpu(irq_stat, cpu).apic_timer_irqs + + per_cpu(irq_stat, cpu).irq0_irqs; +#endif +} + +/* Run after command line and cpu_init init, but before all other checks */ +void nmi_watchdog_default(void) +{ + if (nmi_watchdog != NMI_DEFAULT) + return; +#ifdef CONFIG_X86_64 + nmi_watchdog = NMI_NONE; +#else + if (lapic_watchdog_ok()) + nmi_watchdog = NMI_LOCAL_APIC; + else + nmi_watchdog = NMI_IO_APIC; +#endif +} + +#ifdef CONFIG_SMP +/* + * The performance counters used by NMI_LOCAL_APIC don't trigger when + * the CPU is idle. To make sure the NMI watchdog really ticks on all + * CPUs during the test make them busy. + */ +static __init void nmi_cpu_busy(void *data) +{ + local_irq_enable_in_hardirq(); + /* + * Intentionally don't use cpu_relax here. This is + * to make sure that the performance counter really ticks, + * even if there is a simulator or similar that catches the + * pause instruction. On a real HT machine this is fine because + * all other CPUs are busy with "useless" delay loops and don't + * care if they get somewhat less cycles. + */ + while (endflag == 0) + mb(); +} +#endif + +int __init check_nmi_watchdog(void) +{ + unsigned int *prev_nmi_count; + int cpu; + + if (nmi_watchdog == NMI_NONE || nmi_watchdog == NMI_DISABLED) + return 0; + + if (!atomic_read(&nmi_active)) + return 0; + + prev_nmi_count = kmalloc(NR_CPUS * sizeof(int), GFP_KERNEL); + if (!prev_nmi_count) + goto error; + + printk(KERN_INFO "Testing NMI watchdog ... "); + +#ifdef CONFIG_SMP + if (nmi_watchdog == NMI_LOCAL_APIC) + smp_call_function(nmi_cpu_busy, (void *)&endflag, 0, 0); +#endif + + for_each_possible_cpu(cpu) + prev_nmi_count[cpu] = get_nmi_count(cpu); + local_irq_enable(); + mdelay((20 * 1000) / nmi_hz); /* wait 20 ticks */ + + for_each_online_cpu(cpu) { + if (!per_cpu(wd_enabled, cpu)) + continue; + if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5) { + printk(KERN_WARNING "WARNING: CPU#%d: NMI " + "appears to be stuck (%d->%d)!\n", + cpu, + prev_nmi_count[cpu], + get_nmi_count(cpu)); + per_cpu(wd_enabled, cpu) = 0; + atomic_dec(&nmi_active); + } + } + endflag = 1; + if (!atomic_read(&nmi_active)) { + kfree(prev_nmi_count); + atomic_set(&nmi_active, -1); + goto error; + } + printk("OK.\n"); + + /* + * now that we know it works we can reduce NMI frequency to + * something more reasonable; makes a difference in some configs + */ + if (nmi_watchdog == NMI_LOCAL_APIC) + nmi_hz = lapic_adjust_nmi_hz(1); + + kfree(prev_nmi_count); + return 0; + +error: +#ifdef CONFIG_X86_32 + timer_ack = !cpu_has_tsc; +#endif + return -1; +} + +static int __init setup_nmi_watchdog(char *str) +{ + int nmi; + + if (!strncmp(str, "panic", 5)) { + panic_on_timeout = 1; + str = strchr(str, ','); + if (!str) + return 1; + ++str; + } + + get_option(&str, &nmi); + + if (nmi >= NMI_INVALID || nmi < NMI_NONE) + return 0; + + nmi_watchdog = nmi; + return 1; +} +__setup("nmi_watchdog=", setup_nmi_watchdog); + +/* + * Suspend/resume support + */ +#ifdef CONFIG_PM + +static int nmi_pm_active; /* nmi_active before suspend */ + +static int lapic_nmi_suspend(struct sys_device *dev, pm_message_t state) +{ + /* only CPU0 goes here, other CPUs should be offline */ + nmi_pm_active = atomic_read(&nmi_active); + stop_apic_nmi_watchdog(NULL); + BUG_ON(atomic_read(&nmi_active) != 0); + return 0; +} + +static int lapic_nmi_resume(struct sys_device *dev) +{ + /* only CPU0 goes here, other CPUs should be offline */ + if (nmi_pm_active > 0) { + setup_apic_nmi_watchdog(NULL); + touch_nmi_watchdog(); + } + return 0; +} + +static struct sysdev_class nmi_sysclass = { + .name = "lapic_nmi", + .resume = lapic_nmi_resume, + .suspend = lapic_nmi_suspend, +}; + +static struct sys_device device_lapic_nmi = { + .id = 0, + .cls = &nmi_sysclass, +}; + +static int __init init_lapic_nmi_sysfs(void) +{ + int error; + + /* + * should really be a BUG_ON but b/c this is an + * init call, it just doesn't work. -dcz + */ + if (nmi_watchdog != NMI_LOCAL_APIC) + return 0; + + if (atomic_read(&nmi_active) < 0) + return 0; + + error = sysdev_class_register(&nmi_sysclass); + if (!error) + error = sysdev_register(&device_lapic_nmi); + return error; +} + +/* must come after the local APIC's device_initcall() */ +late_initcall(init_lapic_nmi_sysfs); + +#endif /* CONFIG_PM */ + +static void __acpi_nmi_enable(void *__unused) +{ + apic_write_around(APIC_LVT0, APIC_DM_NMI); +} + +/* + * Enable timer based NMIs on all CPUs: + */ +void acpi_nmi_enable(void) +{ + if (atomic_read(&nmi_active) && nmi_watchdog == NMI_IO_APIC) + on_each_cpu(__acpi_nmi_enable, NULL, 0, 1); +} + +static void __acpi_nmi_disable(void *__unused) +{ + apic_write_around(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); +} + +/* + * Disable timer based NMIs on all CPUs: + */ +void acpi_nmi_disable(void) +{ + if (atomic_read(&nmi_active) && nmi_watchdog == NMI_IO_APIC) + on_each_cpu(__acpi_nmi_disable, NULL, 0, 1); +} + +void setup_apic_nmi_watchdog(void *unused) +{ + if (__get_cpu_var(wd_enabled)) + return; + + /* cheap hack to support suspend/resume */ + /* if cpu0 is not active neither should the other cpus */ + if (smp_processor_id() != 0 && atomic_read(&nmi_active) <= 0) + return; + + switch (nmi_watchdog) { + case NMI_LOCAL_APIC: + /* enable it before to avoid race with handler */ + __get_cpu_var(wd_enabled) = 1; + if (lapic_watchdog_init(nmi_hz) < 0) { + __get_cpu_var(wd_enabled) = 0; + return; + } + /* FALL THROUGH */ + case NMI_IO_APIC: + __get_cpu_var(wd_enabled) = 1; + atomic_inc(&nmi_active); + } +} + +void stop_apic_nmi_watchdog(void *unused) +{ + /* only support LOCAL and IO APICs for now */ + if (nmi_watchdog != NMI_LOCAL_APIC && + nmi_watchdog != NMI_IO_APIC) + return; + if (__get_cpu_var(wd_enabled) == 0) + return; + if (nmi_watchdog == NMI_LOCAL_APIC) + lapic_watchdog_stop(); + __get_cpu_var(wd_enabled) = 0; + atomic_dec(&nmi_active); +} + +/* + * the best way to detect whether a CPU has a 'hard lockup' problem + * is to check it's local APIC timer IRQ counts. If they are not + * changing then that CPU has some problem. + * + * as these watchdog NMI IRQs are generated on every CPU, we only + * have to check the current processor. + * + * since NMIs don't listen to _any_ locks, we have to be extremely + * careful not to rely on unsafe variables. The printk might lock + * up though, so we have to break up any console locks first ... + * [when there will be more tty-related locks, break them up here too!] + */ + +static DEFINE_PER_CPU(unsigned, last_irq_sum); +static DEFINE_PER_CPU(local_t, alert_counter); +static DEFINE_PER_CPU(int, nmi_touch); + +void touch_nmi_watchdog(void) +{ + if (nmi_watchdog > 0) { + unsigned cpu; + + /* + * Tell other CPUs to reset their alert counters. We cannot + * do it ourselves because the alert count increase is not + * atomic. + */ + for_each_present_cpu(cpu) { + if (per_cpu(nmi_touch, cpu) != 1) + per_cpu(nmi_touch, cpu) = 1; + } + } + + /* + * Tickle the softlockup detector too: + */ + touch_softlockup_watchdog(); +} +EXPORT_SYMBOL(touch_nmi_watchdog); + +notrace __kprobes int +nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) +{ + /* + * Since current_thread_info()-> is always on the stack, and we + * always switch the stack NMI-atomically, it's safe to use + * smp_processor_id(). + */ + unsigned int sum; + int touched = 0; + int cpu = smp_processor_id(); + int rc = 0; + + /* check for other users first */ + if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) + == NOTIFY_STOP) { + rc = 1; + touched = 1; + } + + sum = get_timer_irqs(cpu); + + if (__get_cpu_var(nmi_touch)) { + __get_cpu_var(nmi_touch) = 0; + touched = 1; + } + + if (cpu_isset(cpu, backtrace_mask)) { + static DEFINE_SPINLOCK(lock); /* Serialise the printks */ + + spin_lock(&lock); + printk("NMI backtrace for cpu %d\n", cpu); + dump_stack(); + spin_unlock(&lock); + cpu_clear(cpu, backtrace_mask); + } + + /* Could check oops_in_progress here too, but it's safer not to */ + if (mce_in_progress()) + touched = 1; + + /* if the none of the timers isn't firing, this cpu isn't doing much */ + if (!touched && __get_cpu_var(last_irq_sum) == sum) { + /* + * Ayiee, looks like this CPU is stuck ... + * wait a few IRQs (5 seconds) before doing the oops ... + */ + local_inc(&__get_cpu_var(alert_counter)); + if (local_read(&__get_cpu_var(alert_counter)) == 5 * nmi_hz) + /* + * die_nmi will return ONLY if NOTIFY_STOP happens.. + */ + die_nmi("BUG: NMI Watchdog detected LOCKUP", + regs, panic_on_timeout); + } else { + __get_cpu_var(last_irq_sum) = sum; + local_set(&__get_cpu_var(alert_counter), 0); + } + + /* see if the nmi watchdog went off */ + if (!__get_cpu_var(wd_enabled)) + return rc; + switch (nmi_watchdog) { + case NMI_LOCAL_APIC: + rc |= lapic_wd_event(nmi_hz); + break; + case NMI_IO_APIC: + /* + * don't know how to accurately check for this. + * just assume it was a watchdog timer interrupt + * This matches the old behaviour. + */ + rc = 1; + break; + } + return rc; +} + +#ifdef CONFIG_SYSCTL + +static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu) +{ + unsigned char reason = get_nmi_reason(); + char buf[64]; + + sprintf(buf, "NMI received for unknown reason %02x\n", reason); + die_nmi(buf, regs, 1); /* Always panic here */ + return 0; +} + +/* + * proc handler for /proc/sys/kernel/nmi + */ +int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file, + void __user *buffer, size_t *length, loff_t *ppos) +{ + int old_state; + + nmi_watchdog_enabled = (atomic_read(&nmi_active) > 0) ? 1 : 0; + old_state = nmi_watchdog_enabled; + proc_dointvec(table, write, file, buffer, length, ppos); + if (!!old_state == !!nmi_watchdog_enabled) + return 0; + + if (atomic_read(&nmi_active) < 0 || nmi_watchdog == NMI_DISABLED) { + printk(KERN_WARNING + "NMI watchdog is permanently disabled\n"); + return -EIO; + } + + /* if nmi_watchdog is not set yet, then set it */ + nmi_watchdog_default(); + + if (nmi_watchdog == NMI_LOCAL_APIC) { + if (nmi_watchdog_enabled) + enable_lapic_nmi_watchdog(); + else + disable_lapic_nmi_watchdog(); + } else { + printk(KERN_WARNING + "NMI watchdog doesn't know what hardware to touch\n"); + return -EIO; + } + return 0; +} + +#endif /* CONFIG_SYSCTL */ + +int do_nmi_callback(struct pt_regs *regs, int cpu) +{ +#ifdef CONFIG_SYSCTL + if (unknown_nmi_panic) + return unknown_nmi_panic_callback(regs, cpu); +#endif + return 0; +} + +void __trigger_all_cpu_backtrace(void) +{ + int i; + + backtrace_mask = cpu_online_map; + /* Wait for up to 10 seconds for all CPUs to do the backtrace */ + for (i = 0; i < 10 * 1000; i++) { + if (cpus_empty(backtrace_mask)) + break; + mdelay(1); + } +} + +EXPORT_SYMBOL(nmi_active); +EXPORT_SYMBOL(nmi_watchdog); + diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c deleted file mode 100644 index a94dbedf2b74..000000000000 --- a/arch/x86/kernel/nmi_32.c +++ /dev/null @@ -1,506 +0,0 @@ -/* - * NMI watchdog support on APIC systems - * - * Started by Ingo Molnar - * - * Fixes: - * Mikael Pettersson : AMD K7 support for local APIC NMI watchdog. - * Mikael Pettersson : Power Management for local APIC NMI watchdog. - * Mikael Pettersson : Pentium 4 support for local APIC NMI watchdog. - * Pavel Machek and - * Mikael Pettersson : PM converted to driver model. Disable/enable API. - */ - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -#include "mach_traps.h" - -int unknown_nmi_panic; -int nmi_watchdog_enabled; - -static cpumask_t backtrace_mask = CPU_MASK_NONE; - -/* nmi_active: - * >0: the lapic NMI watchdog is active, but can be disabled - * <0: the lapic NMI watchdog has not been set up, and cannot - * be enabled - * 0: the lapic NMI watchdog is disabled, but can be enabled - */ -atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ -static int panic_on_timeout; - -unsigned int nmi_watchdog = NMI_DEFAULT; -static unsigned int nmi_hz = HZ; - -static DEFINE_PER_CPU(short, wd_enabled); - -static int endflag __initdata = 0; - -static inline unsigned int get_nmi_count(int cpu) -{ - return nmi_count(cpu); -} - -static inline int mce_in_progress(void) -{ - return 0; -} - -/* - * Take the local apic timer and PIT/HPET into account. We don't - * know which one is active, when we have highres/dyntick on - */ -static inline unsigned int get_timer_irqs(int cpu) -{ - return per_cpu(irq_stat, cpu).apic_timer_irqs + - per_cpu(irq_stat, cpu).irq0_irqs; -} - -/* Run after command line and cpu_init init, but before all other checks */ -void nmi_watchdog_default(void) -{ - if (nmi_watchdog != NMI_DEFAULT) - return; - if (lapic_watchdog_ok()) - nmi_watchdog = NMI_LOCAL_APIC; - else - nmi_watchdog = NMI_IO_APIC; -} - -#ifdef CONFIG_SMP -/* The performance counters used by NMI_LOCAL_APIC don't trigger when - * the CPU is idle. To make sure the NMI watchdog really ticks on all - * CPUs during the test make them busy. - */ -static __init void nmi_cpu_busy(void *data) -{ - local_irq_enable_in_hardirq(); - /* Intentionally don't use cpu_relax here. This is - to make sure that the performance counter really ticks, - even if there is a simulator or similar that catches the - pause instruction. On a real HT machine this is fine because - all other CPUs are busy with "useless" delay loops and don't - care if they get somewhat less cycles. */ - while (endflag == 0) - mb(); -} -#endif - -int __init check_nmi_watchdog(void) -{ - unsigned int *prev_nmi_count; - int cpu; - - if ((nmi_watchdog == NMI_NONE) || (nmi_watchdog == NMI_DISABLED)) - return 0; - - if (!atomic_read(&nmi_active)) - return 0; - - prev_nmi_count = kmalloc(NR_CPUS * sizeof(int), GFP_KERNEL); - if (!prev_nmi_count) - goto error; - - printk(KERN_INFO "Testing NMI watchdog ... "); - -#ifdef CONFIG_SMP - if (nmi_watchdog == NMI_LOCAL_APIC) - smp_call_function(nmi_cpu_busy, (void *)&endflag, 0, 0); -#endif - - for_each_possible_cpu(cpu) - prev_nmi_count[cpu] = get_nmi_count(cpu); - local_irq_enable(); - mdelay((20*1000)/nmi_hz); // wait 20 ticks - - for_each_online_cpu(cpu) { - if (!per_cpu(wd_enabled, cpu)) - continue; - if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5) { - printk(KERN_WARNING "WARNING: CPU#%d: NMI " - "appears to be stuck (%d->%d)!\n", - cpu, - prev_nmi_count[cpu], - get_nmi_count(cpu)); - per_cpu(wd_enabled, cpu) = 0; - atomic_dec(&nmi_active); - } - } - endflag = 1; - if (!atomic_read(&nmi_active)) { - kfree(prev_nmi_count); - atomic_set(&nmi_active, -1); - goto error; - } - printk("OK.\n"); - - /* now that we know it works we can reduce NMI frequency to - something more reasonable; makes a difference in some configs */ - if (nmi_watchdog == NMI_LOCAL_APIC) - nmi_hz = lapic_adjust_nmi_hz(1); - - kfree(prev_nmi_count); - return 0; -error: - timer_ack = !cpu_has_tsc; - - return -1; -} - -static int __init setup_nmi_watchdog(char *str) -{ - int nmi; - - if (!strncmp(str, "panic", 5)) { - panic_on_timeout = 1; - str = strchr(str, ','); - if (!str) - return 1; - ++str; - } - - get_option(&str, &nmi); - - if ((nmi >= NMI_INVALID) || (nmi < NMI_NONE)) - return 0; - - nmi_watchdog = nmi; - return 1; -} - -__setup("nmi_watchdog=", setup_nmi_watchdog); - - -/* Suspend/resume support */ - -#ifdef CONFIG_PM - -static int nmi_pm_active; /* nmi_active before suspend */ - -static int lapic_nmi_suspend(struct sys_device *dev, pm_message_t state) -{ - /* only CPU0 goes here, other CPUs should be offline */ - nmi_pm_active = atomic_read(&nmi_active); - stop_apic_nmi_watchdog(NULL); - BUG_ON(atomic_read(&nmi_active) != 0); - return 0; -} - -static int lapic_nmi_resume(struct sys_device *dev) -{ - /* only CPU0 goes here, other CPUs should be offline */ - if (nmi_pm_active > 0) { - setup_apic_nmi_watchdog(NULL); - touch_nmi_watchdog(); - } - return 0; -} - - -static struct sysdev_class nmi_sysclass = { - .name = "lapic_nmi", - .resume = lapic_nmi_resume, - .suspend = lapic_nmi_suspend, -}; - -static struct sys_device device_lapic_nmi = { - .id = 0, - .cls = &nmi_sysclass, -}; - -static int __init init_lapic_nmi_sysfs(void) -{ - int error; - - /* should really be a BUG_ON but b/c this is an - * init call, it just doesn't work. -dcz - */ - if (nmi_watchdog != NMI_LOCAL_APIC) - return 0; - - if (atomic_read(&nmi_active) < 0) - return 0; - - error = sysdev_class_register(&nmi_sysclass); - if (!error) - error = sysdev_register(&device_lapic_nmi); - return error; -} -/* must come after the local APIC's device_initcall() */ -late_initcall(init_lapic_nmi_sysfs); - -#endif /* CONFIG_PM */ - -static void __acpi_nmi_enable(void *__unused) -{ - apic_write_around(APIC_LVT0, APIC_DM_NMI); -} - -/* - * Enable timer based NMIs on all CPUs: - */ -void acpi_nmi_enable(void) -{ - if (atomic_read(&nmi_active) && nmi_watchdog == NMI_IO_APIC) - on_each_cpu(__acpi_nmi_enable, NULL, 0, 1); -} - -static void __acpi_nmi_disable(void *__unused) -{ - apic_write_around(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); -} - -/* - * Disable timer based NMIs on all CPUs: - */ -void acpi_nmi_disable(void) -{ - if (atomic_read(&nmi_active) && nmi_watchdog == NMI_IO_APIC) - on_each_cpu(__acpi_nmi_disable, NULL, 0, 1); -} - -void setup_apic_nmi_watchdog(void *unused) -{ - if (__get_cpu_var(wd_enabled)) - return; - - /* cheap hack to support suspend/resume */ - /* if cpu0 is not active neither should the other cpus */ - if ((smp_processor_id() != 0) && (atomic_read(&nmi_active) <= 0)) - return; - - switch (nmi_watchdog) { - case NMI_LOCAL_APIC: - __get_cpu_var(wd_enabled) = 1; /* enable it before to avoid race with handler */ - if (lapic_watchdog_init(nmi_hz) < 0) { - __get_cpu_var(wd_enabled) = 0; - return; - } - /* FALL THROUGH */ - case NMI_IO_APIC: - __get_cpu_var(wd_enabled) = 1; - atomic_inc(&nmi_active); - } -} - -void stop_apic_nmi_watchdog(void *unused) -{ - /* only support LOCAL and IO APICs for now */ - if ((nmi_watchdog != NMI_LOCAL_APIC) && - (nmi_watchdog != NMI_IO_APIC)) - return; - if (__get_cpu_var(wd_enabled) == 0) - return; - if (nmi_watchdog == NMI_LOCAL_APIC) - lapic_watchdog_stop(); - __get_cpu_var(wd_enabled) = 0; - atomic_dec(&nmi_active); -} - -/* - * the best way to detect whether a CPU has a 'hard lockup' problem - * is to check it's local APIC timer IRQ counts. If they are not - * changing then that CPU has some problem. - * - * as these watchdog NMI IRQs are generated on every CPU, we only - * have to check the current processor. - * - * since NMIs don't listen to _any_ locks, we have to be extremely - * careful not to rely on unsafe variables. The printk might lock - * up though, so we have to break up any console locks first ... - * [when there will be more tty-related locks, break them up - * here too!] - */ - -static DEFINE_PER_CPU(unsigned, last_irq_sum); -static DEFINE_PER_CPU(local_t, alert_counter); -static DEFINE_PER_CPU(int, nmi_touch); - -void touch_nmi_watchdog(void) -{ - if (nmi_watchdog > 0) { - unsigned cpu; - - /* - * Tell other CPUs to reset their alert counters. We cannot - * do it ourselves because the alert count increase is not - * atomic. - */ - for_each_present_cpu(cpu) { - if (per_cpu(nmi_touch, cpu) != 1) - per_cpu(nmi_touch, cpu) = 1; - } - } - - /* - * Tickle the softlockup detector too: - */ - touch_softlockup_watchdog(); -} -EXPORT_SYMBOL(touch_nmi_watchdog); - -notrace __kprobes int -nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) -{ - - /* - * Since current_thread_info()-> is always on the stack, and we - * always switch the stack NMI-atomically, it's safe to use - * smp_processor_id(). - */ - unsigned int sum; - int touched = 0; - int cpu = smp_processor_id(); - int rc = 0; - - /* check for other users first */ - if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) - == NOTIFY_STOP) { - rc = 1; - touched = 1; - } - - sum = get_timer_irqs(cpu); - - if (__get_cpu_var(nmi_touch)) { - __get_cpu_var(nmi_touch) = 0; - touched = 1; - } - - if (cpu_isset(cpu, backtrace_mask)) { - static DEFINE_SPINLOCK(lock); /* Serialise the printks */ - - spin_lock(&lock); - printk("NMI backtrace for cpu %d\n", cpu); - dump_stack(); - spin_unlock(&lock); - cpu_clear(cpu, backtrace_mask); - } - - /* Could check oops_in_progress here too, but it's safer not to */ - if (mce_in_progress()) - touched = 1; - - /* if the none of the timers isn't firing, this cpu isn't doing much */ - if (!touched && __get_cpu_var(last_irq_sum) == sum) { - /* - * Ayiee, looks like this CPU is stuck ... - * wait a few IRQs (5 seconds) before doing the oops ... - */ - local_inc(&__get_cpu_var(alert_counter)); - if (local_read(&__get_cpu_var(alert_counter)) == 5*nmi_hz) - /* - * die_nmi will return ONLY if NOTIFY_STOP happens.. - */ - die_nmi("BUG: NMI Watchdog detected LOCKUP", - regs, panic_on_timeout); - } else { - __get_cpu_var(last_irq_sum) = sum; - local_set(&__get_cpu_var(alert_counter), 0); - } - /* see if the nmi watchdog went off */ - if (!__get_cpu_var(wd_enabled)) - return rc; - switch (nmi_watchdog) { - case NMI_LOCAL_APIC: - rc |= lapic_wd_event(nmi_hz); - break; - case NMI_IO_APIC: - /* don't know how to accurately check for this. - * just assume it was a watchdog timer interrupt - * This matches the old behaviour. - */ - rc = 1; - break; - } - return rc; -} - -#ifdef CONFIG_SYSCTL - -static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu) -{ - unsigned char reason = get_nmi_reason(); - char buf[64]; - - sprintf(buf, "NMI received for unknown reason %02x\n", reason); - die_nmi(buf, regs, 1); /* Always panic here */ - return 0; -} - -/* - * proc handler for /proc/sys/kernel/nmi - */ -int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file, - void __user *buffer, size_t *length, loff_t *ppos) -{ - int old_state; - - nmi_watchdog_enabled = (atomic_read(&nmi_active) > 0) ? 1 : 0; - old_state = nmi_watchdog_enabled; - proc_dointvec(table, write, file, buffer, length, ppos); - if (!!old_state == !!nmi_watchdog_enabled) - return 0; - - if (atomic_read(&nmi_active) < 0 || nmi_watchdog == NMI_DISABLED) { - printk( KERN_WARNING "NMI watchdog is permanently disabled\n"); - return -EIO; - } - - /* if nmi_watchdog is not set yet, then set it */ - nmi_watchdog_default(); - - if (nmi_watchdog == NMI_LOCAL_APIC) { - if (nmi_watchdog_enabled) - enable_lapic_nmi_watchdog(); - else - disable_lapic_nmi_watchdog(); - } else { - printk( KERN_WARNING - "NMI watchdog doesn't know what hardware to touch\n"); - return -EIO; - } - return 0; -} - -#endif - -int do_nmi_callback(struct pt_regs *regs, int cpu) -{ -#ifdef CONFIG_SYSCTL - if (unknown_nmi_panic) - return unknown_nmi_panic_callback(regs, cpu); -#endif - return 0; -} - -void __trigger_all_cpu_backtrace(void) -{ - int i; - - backtrace_mask = cpu_online_map; - /* Wait for up to 10 seconds for all CPUs to do the backtrace */ - for (i = 0; i < 10 * 1000; i++) { - if (cpus_empty(backtrace_mask)) - break; - mdelay(1); - } -} - -EXPORT_SYMBOL(nmi_active); -EXPORT_SYMBOL(nmi_watchdog); diff --git a/arch/x86/kernel/nmi_64.c b/arch/x86/kernel/nmi_64.c deleted file mode 100644 index c6802447262c..000000000000 --- a/arch/x86/kernel/nmi_64.c +++ /dev/null @@ -1,477 +0,0 @@ -/* - * NMI watchdog support on APIC systems - * - * Started by Ingo Molnar - * - * Fixes: - * Mikael Pettersson : AMD K7 support for local APIC NMI watchdog. - * Mikael Pettersson : Power Management for local APIC NMI watchdog. - * Pavel Machek and - * Mikael Pettersson : PM converted to driver model. Disable/enable API. - */ - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include -#include - -#include - -int unknown_nmi_panic; -int nmi_watchdog_enabled; - -static cpumask_t backtrace_mask = CPU_MASK_NONE; - -/* nmi_active: - * >0: the lapic NMI watchdog is active, but can be disabled - * <0: the lapic NMI watchdog has not been set up, and cannot - * be enabled - * 0: the lapic NMI watchdog is disabled, but can be enabled - */ -atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ -static int panic_on_timeout; - -unsigned int nmi_watchdog = NMI_DEFAULT; -static unsigned int nmi_hz = HZ; - -static DEFINE_PER_CPU(short, wd_enabled); - -static int endflag __initdata = 0; - -static inline unsigned int get_nmi_count(int cpu) -{ - return cpu_pda(cpu)->__nmi_count; -} - -static inline int mce_in_progress(void) -{ -#ifdef CONFIG_X86_MCE - return atomic_read(&mce_entry) > 0; -#endif - return 0; -} - -/* - * Take the local apic timer and PIT/HPET into account. We don't - * know which one is active, when we have highres/dyntick on - */ -static inline unsigned int get_timer_irqs(int cpu) -{ - return read_pda(apic_timer_irqs) + read_pda(irq0_irqs); -} - -/* Run after command line and cpu_init init, but before all other checks */ -void nmi_watchdog_default(void) -{ - if (nmi_watchdog != NMI_DEFAULT) - return; - nmi_watchdog = NMI_NONE; -} - -#ifdef CONFIG_SMP -/* The performance counters used by NMI_LOCAL_APIC don't trigger when - * the CPU is idle. To make sure the NMI watchdog really ticks on all - * CPUs during the test make them busy. - */ -static __init void nmi_cpu_busy(void *data) -{ - local_irq_enable_in_hardirq(); - /* Intentionally don't use cpu_relax here. This is - to make sure that the performance counter really ticks, - even if there is a simulator or similar that catches the - pause instruction. On a real HT machine this is fine because - all other CPUs are busy with "useless" delay loops and don't - care if they get somewhat less cycles. */ - while (endflag == 0) - mb(); -} -#endif - -int __init check_nmi_watchdog(void) -{ - unsigned int *prev_nmi_count; - int cpu; - - if ((nmi_watchdog == NMI_NONE) || (nmi_watchdog == NMI_DISABLED)) - return 0; - - if (!atomic_read(&nmi_active)) - return 0; - - prev_nmi_count = kmalloc(NR_CPUS * sizeof(int), GFP_KERNEL); - if (!prev_nmi_count) - return -1; - - printk(KERN_INFO "Testing NMI watchdog ... "); - -#ifdef CONFIG_SMP - if (nmi_watchdog == NMI_LOCAL_APIC) - smp_call_function(nmi_cpu_busy, (void *)&endflag, 0, 0); -#endif - - for_each_possible_cpu(cpu) - prev_nmi_count[cpu] = get_nmi_count(cpu); - local_irq_enable(); - mdelay((20*1000)/nmi_hz); // wait 20 ticks - - for_each_online_cpu(cpu) { - if (!per_cpu(wd_enabled, cpu)) - continue; - if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5) { - printk(KERN_WARNING "WARNING: CPU#%d: NMI " - "appears to be stuck (%d->%d)!\n", - cpu, - prev_nmi_count[cpu], - get_nmi_count(cpu)); - per_cpu(wd_enabled, cpu) = 0; - atomic_dec(&nmi_active); - } - } - endflag = 1; - if (!atomic_read(&nmi_active)) { - kfree(prev_nmi_count); - atomic_set(&nmi_active, -1); - return -1; - } - printk("OK.\n"); - - /* now that we know it works we can reduce NMI frequency to - something more reasonable; makes a difference in some configs */ - if (nmi_watchdog == NMI_LOCAL_APIC) - nmi_hz = lapic_adjust_nmi_hz(1); - - kfree(prev_nmi_count); - return 0; -} - -static int __init setup_nmi_watchdog(char *str) -{ - int nmi; - - if (!strncmp(str,"panic",5)) { - panic_on_timeout = 1; - str = strchr(str, ','); - if (!str) - return 1; - ++str; - } - - get_option(&str, &nmi); - - if ((nmi >= NMI_INVALID) || (nmi < NMI_NONE)) - return 0; - - nmi_watchdog = nmi; - return 1; -} - -__setup("nmi_watchdog=", setup_nmi_watchdog); - -#ifdef CONFIG_PM - -static int nmi_pm_active; /* nmi_active before suspend */ - -static int lapic_nmi_suspend(struct sys_device *dev, pm_message_t state) -{ - /* only CPU0 goes here, other CPUs should be offline */ - nmi_pm_active = atomic_read(&nmi_active); - stop_apic_nmi_watchdog(NULL); - BUG_ON(atomic_read(&nmi_active) != 0); - return 0; -} - -static int lapic_nmi_resume(struct sys_device *dev) -{ - /* only CPU0 goes here, other CPUs should be offline */ - if (nmi_pm_active > 0) { - setup_apic_nmi_watchdog(NULL); - touch_nmi_watchdog(); - } - return 0; -} - -static struct sysdev_class nmi_sysclass = { - .name = "lapic_nmi", - .resume = lapic_nmi_resume, - .suspend = lapic_nmi_suspend, -}; - -static struct sys_device device_lapic_nmi = { - .id = 0, - .cls = &nmi_sysclass, -}; - -static int __init init_lapic_nmi_sysfs(void) -{ - int error; - - /* should really be a BUG_ON but b/c this is an - * init call, it just doesn't work. -dcz - */ - if (nmi_watchdog != NMI_LOCAL_APIC) - return 0; - - if (atomic_read(&nmi_active) < 0) - return 0; - - error = sysdev_class_register(&nmi_sysclass); - if (!error) - error = sysdev_register(&device_lapic_nmi); - return error; -} -/* must come after the local APIC's device_initcall() */ -late_initcall(init_lapic_nmi_sysfs); - -#endif /* CONFIG_PM */ - -static void __acpi_nmi_enable(void *__unused) -{ - apic_write_around(APIC_LVT0, APIC_DM_NMI); -} - -/* - * Enable timer based NMIs on all CPUs: - */ -void acpi_nmi_enable(void) -{ - if (atomic_read(&nmi_active) && nmi_watchdog == NMI_IO_APIC) - on_each_cpu(__acpi_nmi_enable, NULL, 0, 1); -} - -static void __acpi_nmi_disable(void *__unused) -{ - apic_write_around(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); -} - -/* - * Disable timer based NMIs on all CPUs: - */ -void acpi_nmi_disable(void) -{ - if (atomic_read(&nmi_active) && nmi_watchdog == NMI_IO_APIC) - on_each_cpu(__acpi_nmi_disable, NULL, 0, 1); -} - -void setup_apic_nmi_watchdog(void *unused) -{ - if (__get_cpu_var(wd_enabled)) - return; - - /* cheap hack to support suspend/resume */ - /* if cpu0 is not active neither should the other cpus */ - if ((smp_processor_id() != 0) && (atomic_read(&nmi_active) <= 0)) - return; - - switch (nmi_watchdog) { - case NMI_LOCAL_APIC: - __get_cpu_var(wd_enabled) = 1; - if (lapic_watchdog_init(nmi_hz) < 0) { - __get_cpu_var(wd_enabled) = 0; - return; - } - /* FALL THROUGH */ - case NMI_IO_APIC: - __get_cpu_var(wd_enabled) = 1; - atomic_inc(&nmi_active); - } -} - -void stop_apic_nmi_watchdog(void *unused) -{ - /* only support LOCAL and IO APICs for now */ - if ((nmi_watchdog != NMI_LOCAL_APIC) && - (nmi_watchdog != NMI_IO_APIC)) - return; - if (__get_cpu_var(wd_enabled) == 0) - return; - if (nmi_watchdog == NMI_LOCAL_APIC) - lapic_watchdog_stop(); - __get_cpu_var(wd_enabled) = 0; - atomic_dec(&nmi_active); -} - -/* - * the best way to detect whether a CPU has a 'hard lockup' problem - * is to check it's local APIC timer IRQ counts. If they are not - * changing then that CPU has some problem. - * - * as these watchdog NMI IRQs are generated on every CPU, we only - * have to check the current processor. - */ - -static DEFINE_PER_CPU(unsigned, last_irq_sum); -static DEFINE_PER_CPU(local_t, alert_counter); -static DEFINE_PER_CPU(int, nmi_touch); - -void touch_nmi_watchdog(void) -{ - if (nmi_watchdog > 0) { - unsigned cpu; - - /* - * Tell other CPUs to reset their alert counters. We cannot - * do it ourselves because the alert count increase is not - * atomic. - */ - for_each_present_cpu(cpu) { - if (per_cpu(nmi_touch, cpu) != 1) - per_cpu(nmi_touch, cpu) = 1; - } - } - - touch_softlockup_watchdog(); -} -EXPORT_SYMBOL(touch_nmi_watchdog); - -notrace __kprobes int -nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) -{ - unsigned int sum; - int touched = 0; - int cpu = smp_processor_id(); - int rc = 0; - - /* check for other users first */ - if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) - == NOTIFY_STOP) { - rc = 1; - touched = 1; - } - - sum = get_timer_irqs(cpu); - - if (__get_cpu_var(nmi_touch)) { - __get_cpu_var(nmi_touch) = 0; - touched = 1; - } - - if (cpu_isset(cpu, backtrace_mask)) { - static DEFINE_SPINLOCK(lock); /* Serialise the printks */ - - spin_lock(&lock); - printk("NMI backtrace for cpu %d\n", cpu); - dump_stack(); - spin_unlock(&lock); - cpu_clear(cpu, backtrace_mask); - } - - if (mce_in_progress()) - touched = 1; - - /* if the apic timer isn't firing, this cpu isn't doing much */ - if (!touched && __get_cpu_var(last_irq_sum) == sum) { - /* - * Ayiee, looks like this CPU is stuck ... - * wait a few IRQs (5 seconds) before doing the oops ... - */ - local_inc(&__get_cpu_var(alert_counter)); - if (local_read(&__get_cpu_var(alert_counter)) == 5*nmi_hz) - die_nmi("NMI Watchdog detected LOCKUP", - regs, panic_on_timeout); - } else { - __get_cpu_var(last_irq_sum) = sum; - local_set(&__get_cpu_var(alert_counter), 0); - } - - /* see if the nmi watchdog went off */ - if (!__get_cpu_var(wd_enabled)) - return rc; - switch (nmi_watchdog) { - case NMI_LOCAL_APIC: - rc |= lapic_wd_event(nmi_hz); - break; - case NMI_IO_APIC: - /* don't know how to accurately check for this. - * just assume it was a watchdog timer interrupt - * This matches the old behaviour. - */ - rc = 1; - break; - } - return rc; -} - -#ifdef CONFIG_SYSCTL - -static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu) -{ - unsigned char reason = get_nmi_reason(); - char buf[64]; - - sprintf(buf, "NMI received for unknown reason %02x\n", reason); - die_nmi(buf, regs, 1); /* Always panic here */ - return 0; -} - -/* - * proc handler for /proc/sys/kernel/nmi - */ -int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file, - void __user *buffer, size_t *length, loff_t *ppos) -{ - int old_state; - - nmi_watchdog_enabled = (atomic_read(&nmi_active) > 0) ? 1 : 0; - old_state = nmi_watchdog_enabled; - proc_dointvec(table, write, file, buffer, length, ppos); - if (!!old_state == !!nmi_watchdog_enabled) - return 0; - - if (atomic_read(&nmi_active) < 0 || nmi_watchdog == NMI_DISABLED) { - printk( KERN_WARNING "NMI watchdog is permanently disabled\n"); - return -EIO; - } - - /* if nmi_watchdog is not set yet, then set it */ - nmi_watchdog_default(); - - if (nmi_watchdog == NMI_LOCAL_APIC) { - if (nmi_watchdog_enabled) - enable_lapic_nmi_watchdog(); - else - disable_lapic_nmi_watchdog(); - } else { - printk( KERN_WARNING - "NMI watchdog doesn't know what hardware to touch\n"); - return -EIO; - } - return 0; -} - -#endif - -int do_nmi_callback(struct pt_regs *regs, int cpu) -{ -#ifdef CONFIG_SYSCTL - if (unknown_nmi_panic) - return unknown_nmi_panic_callback(regs, cpu); -#endif - return 0; -} - -void __trigger_all_cpu_backtrace(void) -{ - int i; - - backtrace_mask = cpu_online_map; - /* Wait for up to 10 seconds for all CPUs to do the backtrace */ - for (i = 0; i < 10 * 1000; i++) { - if (cpus_empty(backtrace_mask)) - break; - mdelay(1); - } -} - -EXPORT_SYMBOL(nmi_active); -EXPORT_SYMBOL(nmi_watchdog); -- cgit v1.2.3 From 19ec673ced067316b9732bc6d1c4ff4052e5f795 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Wed, 28 May 2008 23:00:47 +0400 Subject: x86: nmi - fix incorrect NMI watchdog used by default The commit commit 4b82b277707a39b97271439c475f186f63ec4692 Author: Cyrill Gorcunov Date: Sat May 24 19:36:35 2008 +0400 set nmi_watchdog to NMI_IO_APIC as by default. This causes hangs on some machines with buggy watchdogs. Fix it - i.e. restore old behaviour. Thanks to Sitsofe Wheeler and Adrian Bunk for catching the problem and Maciej W. Rozycki for explanation what is going on there. Signed-off-by: Cyrill Gorcunov CC: Maciej W. Rozycki Signed-off-by: Ingo Molnar --- arch/x86/kernel/nmi.c | 16 +++++++++------- include/asm-x86/nmi.h | 4 +++- 2 files changed, 12 insertions(+), 8 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index 69a839fc1eb0..3671a9f3564b 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -84,20 +84,15 @@ static inline unsigned int get_timer_irqs(int cpu) #endif } +#ifdef CONFIG_X86_64 /* Run after command line and cpu_init init, but before all other checks */ void nmi_watchdog_default(void) { if (nmi_watchdog != NMI_DEFAULT) return; -#ifdef CONFIG_X86_64 nmi_watchdog = NMI_NONE; -#else - if (lapic_watchdog_ok()) - nmi_watchdog = NMI_LOCAL_APIC; - else - nmi_watchdog = NMI_IO_APIC; -#endif } +#endif #ifdef CONFIG_SMP /* @@ -488,8 +483,15 @@ int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file, return -EIO; } +#ifdef CONFIG_X86_64 /* if nmi_watchdog is not set yet, then set it */ nmi_watchdog_default(); +#else + if (lapic_watchdog_ok()) + nmi_watchdog = NMI_LOCAL_APIC; + else + nmi_watchdog = NMI_IO_APIC; +#endif if (nmi_watchdog == NMI_LOCAL_APIC) { if (nmi_watchdog_enabled) diff --git a/include/asm-x86/nmi.h b/include/asm-x86/nmi.h index 1e8f34d7ab68..6f4d44fc051f 100644 --- a/include/asm-x86/nmi.h +++ b/include/asm-x86/nmi.h @@ -38,9 +38,11 @@ static inline void unset_nmi_pm_callback(struct pm_dev *dev) #ifdef CONFIG_X86_64 extern void default_do_nmi(struct pt_regs *); +extern void nmi_watchdog_default(void); +#else +#define nmi_watchdog_default(void) do {} while (0) #endif -extern void nmi_watchdog_default(void); extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); extern int check_nmi_watchdog(void); extern int nmi_watchdog_enabled; -- cgit v1.2.3 From 88ff0a474e98f869d8c321e29481f298320100d6 Mon Sep 17 00:00:00 2001 From: Hiroshi Shimamoto Date: Tue, 27 May 2008 18:49:39 -0700 Subject: x86: coding style fixes for nmi.c before total: 1 errors, 6 warnings, 534 lines checked after total: 0 errors, 1 warnings, 532 lines checked Signed-off-by: Hiroshi Shimamoto Signed-off-by: Ingo Molnar --- arch/x86/kernel/nmi.c | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index 3671a9f3564b..bf143f191fb1 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -23,9 +23,8 @@ #include #include #include +#include -#include -#include #include #include @@ -45,13 +44,16 @@ static cpumask_t backtrace_mask = CPU_MASK_NONE; * 0: the lapic NMI watchdog is disabled, but can be enabled */ atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ -static int panic_on_timeout; +EXPORT_SYMBOL(nmi_active); unsigned int nmi_watchdog = NMI_DEFAULT; +EXPORT_SYMBOL(nmi_watchdog); + +static int panic_on_timeout; static unsigned int nmi_hz = HZ; static DEFINE_PER_CPU(short, wd_enabled); -static int endflag __initdata = 0; +static int endflag __initdata; static inline unsigned int get_nmi_count(int cpu) { @@ -404,7 +406,7 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) static DEFINE_SPINLOCK(lock); /* Serialise the printks */ spin_lock(&lock); - printk("NMI backtrace for cpu %d\n", cpu); + printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu); dump_stack(); spin_unlock(&lock); cpu_clear(cpu, backtrace_mask); @@ -529,7 +531,3 @@ void __trigger_all_cpu_backtrace(void) mdelay(1); } } - -EXPORT_SYMBOL(nmi_active); -EXPORT_SYMBOL(nmi_watchdog); - -- cgit v1.2.3 From 3d1ba1da2b4ff4ace7801e99fb9a3095b182d847 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 2 Jun 2008 13:50:10 +0200 Subject: x86: fix nmi.c build bug apic.h needs to be included for the apic_write_around() definition. --- arch/x86/kernel/nmi.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index bf143f191fb1..cbd4fa3c475b 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -11,6 +11,8 @@ * Mikael Pettersson : PM converted to driver model. Disable/enable API. */ +#include + #include #include #include -- cgit v1.2.3 From 3ed3f06295e69700fa808396f7b350bff2b69de0 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Wed, 4 Jun 2008 01:00:47 +0400 Subject: x86: nmi - consolidate nmi_watchdog_default for 32bit mode 64bit mode bootstrap code does set nmi_watchdog to NMI_NONE by default and doing the same on 32bit mode is safe too. Such an action saves us from several #ifdef. Btw, my previous commit commit 19ec673ced067316b9732bc6d1c4ff4052e5f795 Author: Cyrill Gorcunov Date: Wed May 28 23:00:47 2008 +0400 x86: nmi - fix incorrect NMI watchdog used by default did not fix the problem completely, moreover it introduced additional bug - nmi_watchdog would be set to either NMI_LOCAL_APIC or NMI_IO_APIC _regardless_ to boot option if being enabled thru /proc/sys/kernel/nmi_watchdog. Sorry for that. Fix it too. Signed-off-by: Cyrill Gorcunov Cc: mingo@redhat.com Cc: hpa@zytor.com Cc: macro@linux-mips.org Signed-off-by: Thomas Gleixner --- arch/x86/kernel/nmi.c | 14 ++++++++------ include/asm-x86/nmi.h | 4 +--- 2 files changed, 9 insertions(+), 9 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index cbd4fa3c475b..27ca8f69b466 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -487,14 +487,16 @@ int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file, return -EIO; } -#ifdef CONFIG_X86_64 /* if nmi_watchdog is not set yet, then set it */ nmi_watchdog_default(); -#else - if (lapic_watchdog_ok()) - nmi_watchdog = NMI_LOCAL_APIC; - else - nmi_watchdog = NMI_IO_APIC; + +#ifdef CONFIG_X86_32 + if (nmi_watchdog == NMI_NONE) { + if (lapic_watchdog_ok()) + nmi_watchdog = NMI_LOCAL_APIC; + else + nmi_watchdog = NMI_IO_APIC; + } #endif if (nmi_watchdog == NMI_LOCAL_APIC) { diff --git a/include/asm-x86/nmi.h b/include/asm-x86/nmi.h index 972a4f6f799a..470bb4aacb75 100644 --- a/include/asm-x86/nmi.h +++ b/include/asm-x86/nmi.h @@ -38,12 +38,10 @@ static inline void unset_nmi_pm_callback(struct pm_dev *dev) #ifdef CONFIG_X86_64 extern void default_do_nmi(struct pt_regs *); -extern void nmi_watchdog_default(void); -#else -#define nmi_watchdog_default() do { } while (0) #endif extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); +extern void nmi_watchdog_default(void); extern int check_nmi_watchdog(void); extern int nmi_watchdog_enabled; extern int avail_to_resrv_perfctr_nmi_bit(unsigned int); -- cgit v1.2.3 From 1a1b1d1322ebd1ece405f3057cdd408bc77e391d Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Wed, 4 Jun 2008 01:00:58 +0400 Subject: x86: watchdog - check for CPU is being supported This patch does check if CPU is being recongnized before call the unreserve(). Since enable_lapic_nmi_watchdog() does have such a check the same is make sense here too in a sake of code consistency (but nothing more). Signed-off-by: Cyrill Gorcunov Cc: mingo@redhat.com Cc: hpa@zytor.com Cc: macro@linux-mips.org Signed-off-by: Thomas Gleixner --- arch/x86/kernel/cpu/perfctr-watchdog.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c index f9ae93adffe5..ddda4b64f545 100644 --- a/arch/x86/kernel/cpu/perfctr-watchdog.c +++ b/arch/x86/kernel/cpu/perfctr-watchdog.c @@ -181,7 +181,9 @@ void disable_lapic_nmi_watchdog(void) return; on_each_cpu(stop_apic_nmi_watchdog, NULL, 0, 1); - wd_ops->unreserve(); + + if (wd_ops) + wd_ops->unreserve(); BUG_ON(atomic_read(&nmi_active) != 0); } -- cgit v1.2.3 From 75b9f5d2a0318da9d5e694a9a1be33f46b4c021e Mon Sep 17 00:00:00 2001 From: "mingo@elte.hu" Date: Thu, 5 Jun 2008 11:18:12 +0200 Subject: x86, nmi: fix build fix: arch/x86/kernel/built-in.o: In function `proc_nmi_enabled': : undefined reference to `nmi_watchdog_default' arch/x86/kernel/built-in.o: In function `native_smp_prepare_cpus': : undefined reference to `nmi_watchdog_default' Signed-off-by: Ingo Molnar --- arch/x86/kernel/nmi.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index 27ca8f69b466..19f1b95265cf 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -88,7 +88,6 @@ static inline unsigned int get_timer_irqs(int cpu) #endif } -#ifdef CONFIG_X86_64 /* Run after command line and cpu_init init, but before all other checks */ void nmi_watchdog_default(void) { @@ -96,7 +95,6 @@ void nmi_watchdog_default(void) return; nmi_watchdog = NMI_NONE; } -#endif #ifdef CONFIG_SMP /* -- cgit v1.2.3 From f781b03c4b1c713ac000877c8bbc31fc4164a29b Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Thu, 12 Jun 2008 17:08:16 +0400 Subject: x86: touch_nmi_watchdog(): reset alert counters for supported nmi_watchdog modes only The checking 'if nmi_watchdog > 0' (ie NMI_NONE) is quite fast but it has a side effect - it's taken even if nmi_watchdog = NMI_DISABLED. Nowadays nmi_watchdog is set up to NMI_NONE by default so this condition is properly taken most the time but we better show this explicitly. Signed-off-by: Cyrill Gorcunov Signed-off-by: Ingo Molnar --- arch/x86/kernel/nmi.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index 19f1b95265cf..326a8f4f50f8 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -354,7 +354,8 @@ static DEFINE_PER_CPU(int, nmi_touch); void touch_nmi_watchdog(void) { - if (nmi_watchdog > 0) { + if (nmi_watchdog == NMI_LOCAL_APIC || + nmi_watchdog == NMI_IO_APIC) { unsigned cpu; /* -- cgit v1.2.3 From b8e0418b2a25f975c3b8764030c24b7253d33a68 Mon Sep 17 00:00:00 2001 From: Glauber Costa Date: Mon, 16 Jun 2008 19:59:08 -0300 Subject: x86: fix typo CONFIX -> CONFIG Signed-off-by: Ingo Molnar --- arch/x86/kernel/nmi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index 326a8f4f50f8..fd680c73ba77 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -68,7 +68,7 @@ static inline unsigned int get_nmi_count(int cpu) static inline int mce_in_progress(void) { -#if defined(CONFIX_X86_64) && defined(CONFIG_X86_MCE) +#if defined(CONFIG_X86_64) && defined(CONFIG_X86_MCE) return atomic_read(&mce_entry) > 0; #endif return 0; -- cgit v1.2.3