* [RFC] x86: merge nmi_32-64 to nmi.c @ 2008-05-17 19:22 Cyrill Gorcunov 2008-05-17 20:28 ` Tom Spink 0 siblings, 1 reply; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-17 19:22 UTC (permalink / raw) To: Ingo Molnar, H. Peter Anvin, Thomas Gleixner; +Cc: LKML, Jiri Slaby, Andi Kleen This is an attempt to merge nmi_32.c and nmi_64.c to a single file since they are match about 75%. I'm absolutely sure there are some nits or even bugs so I need help from community by a strong review. Any feedback is appreciated. To make review process a bit easier here is a list of what was changed in files by merging procedure: - a few helper functions added (placed on top part of nmi.c) - nmi_watchdog_default() now handles both 32/64 modes - check_nmi_watchdog() uses for_each_online_cpu() even in 32bit mode since cpu_online_map is set up after cpu_callin_map anyway - last_irq_sums and alert_counter was defined as static arrays in 32bit mode, so they were changed to per_cpu variables - and the most notable change is that I've changed logic a bit for procedures touch_nmi_watchdog() and nmi_watchdog_tick() used in 32bit mode (to be familiar with 64bit mode). - die_nmi() in traps_64.c now prints almost the same message as it was for traps_32.c so we don't need to specify different args on die_nmi() - die_nmi() definition is moved out of nmi.c to the header (nmi.h) So please get this RFC a strong review. At least we always able to just drop it ;) Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com> --- arch/x86/kernel/Makefile | 2 +- arch/x86/kernel/nmi.c | 578 ++++++++++++++++++++++++++++++++++++++++++++ arch/x86/kernel/nmi_32.c | 472 ------------------------------------ arch/x86/kernel/nmi_64.c | 482 ------------------------------------ arch/x86/kernel/traps_64.c | 4 +- include/asm-x86/nmi.h | 2 +- 6 files changed, 583 insertions(+), 957 deletions(-) diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 5e618c3..d5dd666 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 0000000..8cc6491 --- /dev/null +++ b/arch/x86/kernel/nmi.c @@ -0,0 +1,578 @@ +/* + * NMI watchdog support on APIC systems + * + * Started by Ingo Molnar <mingo@redhat.com> + * + * 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 <linux/nmi.h> +#include <linux/mm.h> +#include <linux/delay.h> +#include <linux/interrupt.h> +#include <linux/module.h> +#include <linux/sysdev.h> +#include <linux/sysctl.h> +#include <linux/kprobes.h> +#include <linux/cpumask.h> +#include <linux/kdebug.h> +#include <linux/kernel_stat.h> + +#include <asm/smp.h> +#include <asm/nmi.h> +#include <asm/proto.h> +#include <asm/mce.h> +#include <asm/timer.h> + +#include <mach_traps.h> + +int unknown_nmi_panic; +int nmi_watchdog_enabled; + +#ifdef CONFIG_X86_64 +int panic_on_unrecovered_nmi; +#endif +static int panic_on_timeout; + +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 */ +unsigned int nmi_watchdog = NMI_DEFAULT; + +static DEFINE_PER_CPU(short, wd_enabled); +static unsigned int nmi_hz = HZ; +static int endflag __initdata = 0; + +/* a few helper functions */ +#ifdef CONFIG_X86_64 + +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; +} + +static inline void __die_nmi(char *str, struct pt_regs *regs, int do_panic) +{ + die_nmi(str, regs, do_panic); +} + +#else /* !CONFIG_X86_64 */ + +static inline unsigned int get_nmi_count(int cpu) +{ + return nmi_count(cpu); +} + +static inline int mce_in_progress(void) +{ + return 0; +} + +static inline void __die_nmi(char *str, struct pt_regs *regs, int do_panic) +{ + die_nmi(regs, str); +} + +#endif /* !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 +} + +#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 /* CONFIG_SMP */ + +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 for 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; + +#ifdef CONFIG_X86_64 + if (!strncmp(str,"panic",5)) { + panic_on_timeout = 1; + str = strchr(str, ','); + if (!str) + return 1; + ++str; + } +#endif + + 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) +{ +#ifdef CONFIG_X86_64 + apic_write(APIC_LVT0, APIC_DM_NMI); +#else + apic_write_around(APIC_LVT0, APIC_DM_NMI); +#endif +} + +/* 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(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. + * + * 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; + } + + /* + * Take the local apic timer and PIT/HPET into account. We don't + * know which one is active, when we have highres/dyntick on + */ +#ifdef CONFIG_X86_64 + sum = read_pda(apic_timer_irqs) + read_pda(irq0_irqs); +#else + sum = per_cpu(irq_stat, cpu).apic_timer_irqs + + per_cpu(irq_stat, cpu).irq0_irqs; +#endif + + if (__get_cpu_var(nmi_touch)) { + __get_cpu_var(nmi_touch) = 0; + touched = 1; + } + + if (cpu_isset(cpu, backtrace_mask)) { + static DEFINE_SPINLOCK(lock); + 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 do + */ + 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("NMI Watchdog detected LOCKUP\n", + 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_X86_64 +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(); +} +#endif /* CONFIG_X86_64 */ + +#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); + 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 11b14bb..0000000 --- a/arch/x86/kernel/nmi_32.c +++ /dev/null @@ -1,472 +0,0 @@ -/* - * NMI watchdog support on APIC systems - * - * Started by Ingo Molnar <mingo@redhat.com> - * - * 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 <linux/delay.h> -#include <linux/interrupt.h> -#include <linux/module.h> -#include <linux/nmi.h> -#include <linux/sysdev.h> -#include <linux/sysctl.h> -#include <linux/percpu.h> -#include <linux/kprobes.h> -#include <linux/cpumask.h> -#include <linux/kernel_stat.h> -#include <linux/kdebug.h> -#include <linux/slab.h> - -#include <asm/smp.h> -#include <asm/nmi.h> -#include <asm/timer.h> - -#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 */ - -unsigned int nmi_watchdog = NMI_DEFAULT; -static unsigned int nmi_hz = HZ; - -static DEFINE_PER_CPU(short, wd_enabled); - -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 - * 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] = nmi_count(cpu); - 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 - if (!per_cpu(wd_enabled, cpu)) - continue; - if (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)); - 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; - - 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(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 unsigned int - last_irq_sums [NR_CPUS], - alert_counter [NR_CPUS]; - -void touch_nmi_watchdog(void) -{ - if (nmi_watchdog > 0) { - unsigned cpu; - - /* - * Just reset the alert counters, (other CPUs might be - * spinning on locks we hold): - */ - for_each_present_cpu(cpu) { - if (alert_counter[cpu]) - alert_counter[cpu] = 0; - } - } - - /* - * Tickle the softlockup detector too: - */ - touch_softlockup_watchdog(); -} -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) -{ - - /* - * 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; - } - - 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); - } - - /* - * 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 the none of the timers isn't firing, this cpu isn't doing much */ - if (!touched && last_irq_sums[cpu] == 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) - /* - * 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; - } - /* 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(regs, buf); - 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 == NMI_DEFAULT) { - if (lapic_watchdog_ok()) - nmi_watchdog = NMI_LOCAL_APIC; - else - nmi_watchdog = NMI_IO_APIC; - } - - 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 5a29ded..0000000 --- a/arch/x86/kernel/nmi_64.c +++ /dev/null @@ -1,482 +0,0 @@ -/* - * NMI watchdog support on APIC systems - * - * Started by Ingo Molnar <mingo@redhat.com> - * - * 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 <linux/nmi.h> -#include <linux/mm.h> -#include <linux/delay.h> -#include <linux/interrupt.h> -#include <linux/module.h> -#include <linux/sysdev.h> -#include <linux/sysctl.h> -#include <linux/kprobes.h> -#include <linux/cpumask.h> -#include <linux/kdebug.h> - -#include <asm/smp.h> -#include <asm/nmi.h> -#include <asm/proto.h> -#include <asm/mce.h> - -#include <mach_traps.h> - -int unknown_nmi_panic; -int nmi_watchdog_enabled; -int panic_on_unrecovered_nmi; - -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); - -/* 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; -} - -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 - * 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) -{ - 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 (cpu = 0; cpu < NR_CPUS; cpu++) - prev_nmi_count[cpu] = cpu_pda(cpu)->__nmi_count; - 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) { - printk(KERN_WARNING "WARNING: CPU#%d: NMI " - "appears to be stuck (%d->%d)!\n", - cpu, - prev_nmi_count[cpu], - cpu_pda(cpu)->__nmi_count); - 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(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(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) -{ - 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 = read_pda(apic_timer_irqs) + read_pda(irq0_irqs); - 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); - } - -#ifdef CONFIG_X86_MCE - /* Could check oops_in_progress here too, but it's safer - not too */ - if (atomic_read(&mce_entry) > 0) - 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) { - /* - * 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 on CPU %d\n", 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; -} - -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) -{ - 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/traps_64.c b/arch/x86/kernel/traps_64.c index adff76e..6a048ce 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); diff --git a/include/asm-x86/nmi.h b/include/asm-x86/nmi.h index 1e36302..69a19b3 100644 --- a/include/asm-x86/nmi.h +++ b/include/asm-x86/nmi.h @@ -41,7 +41,7 @@ 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) +extern void die_nmi(struct pt_regs *, const char *msg); #endif extern int check_nmi_watchdog(void); ^ permalink raw reply related [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 19:22 [RFC] x86: merge nmi_32-64 to nmi.c Cyrill Gorcunov @ 2008-05-17 20:28 ` Tom Spink 2008-05-17 20:52 ` Maciej W. Rozycki 0 siblings, 1 reply; 35+ messages in thread From: Tom Spink @ 2008-05-17 20:28 UTC (permalink / raw) To: Cyrill Gorcunov Cc: Ingo Molnar, H. Peter Anvin, Thomas Gleixner, LKML, Jiri Slaby, Andi Kleen 2008/5/17 Cyrill Gorcunov <gorcunov@gmail.com>: > +/* a few helper functions */ > +#ifdef CONFIG_X86_64 > + > +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; > +} > + > +static inline void __die_nmi(char *str, struct pt_regs *regs, int do_panic) > +{ > + die_nmi(str, regs, do_panic); > +} > + > +#else /* !CONFIG_X86_64 */ > + > +static inline unsigned int get_nmi_count(int cpu) > +{ > + return nmi_count(cpu); > +} > + > +static inline int mce_in_progress(void) > +{ > + return 0; > +} > + > +static inline void __die_nmi(char *str, struct pt_regs *regs, int do_panic) > +{ > + die_nmi(regs, str); > +} > + > +#endif /* !CONFIG_X86_64 */ Hi, I've always wondered if it's cleaner to define variants of functions like this with the conditionals inside the function, as opposed to one big conditional encapsulating all these functions. IMO, it's cleaner to define the function with conditionals to define it's particular behaviour in the two different cases, because that way there is one definition of the function with both different behaviours inside, e.g.: 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 } I know it introduces a lot of these conditionals, but at least there is one place to look for the get_nmi_count function, instead of searching for all variants of the function. Just a thought! -- Regards, Tom Spink ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 20:28 ` Tom Spink @ 2008-05-17 20:52 ` Maciej W. Rozycki 2008-05-17 21:40 ` Thomas Gleixner 2008-05-17 21:48 ` Mikael Pettersson 0 siblings, 2 replies; 35+ messages in thread From: Maciej W. Rozycki @ 2008-05-17 20:52 UTC (permalink / raw) To: Tom Spink Cc: Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, Thomas Gleixner, LKML, Jiri Slaby, Andi Kleen On Sat, 17 May 2008, Tom Spink wrote: > 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 > } > > I know it introduces a lot of these conditionals, but at least there > is one place to look for the get_nmi_count function, instead of > searching for all variants of the function. Well, I suppose some header should provide a definition like: #ifdef CONFIG_X86_64 #define cpu_x86_64 1 #else #define cpu_x86_64 0 #endif and the you can remove the horrible #ifdef clutter and make the quoted function look like: static inline unsigned int get_nmi_count(int cpu) { return cpu_x86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); } Much better -- isn't it? Maciej ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 20:52 ` Maciej W. Rozycki @ 2008-05-17 21:40 ` Thomas Gleixner 2008-05-18 7:25 ` Jeremy Fitzhardinge 2008-05-18 10:15 ` Andi Kleen 2008-05-17 21:48 ` Mikael Pettersson 1 sibling, 2 replies; 35+ messages in thread From: Thomas Gleixner @ 2008-05-17 21:40 UTC (permalink / raw) To: Maciej W. Rozycki Cc: Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen On Sat, 17 May 2008, Maciej W. Rozycki wrote: > On Sat, 17 May 2008, Tom Spink wrote: > > > 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 > > } > > > > I know it introduces a lot of these conditionals, but at least there > > is one place to look for the get_nmi_count function, instead of > > searching for all variants of the function. > > Well, I suppose some header should provide a definition like: > > #ifdef CONFIG_X86_64 > #define cpu_x86_64 1 > #else > #define cpu_x86_64 0 > #endif > > and the you can remove the horrible #ifdef clutter and make the quoted > function look like: > > static inline unsigned int get_nmi_count(int cpu) > { > return cpu_x86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); > } > > Much better -- isn't it? Definitely, but we should do it at the Kconfig level which allows us to have integer defines as well, so we end up with something like: static inline unsigned int get_nmi_count(int cpu) { return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); } That way we do not even have to think about which header to select for the include and the association of the selector is stronger than the cpu_x86_64 one, isn't it ? Thanks, tglx ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 21:40 ` Thomas Gleixner @ 2008-05-18 7:25 ` Jeremy Fitzhardinge 2008-05-18 7:38 ` Cyrill Gorcunov ` (2 more replies) 2008-05-18 10:15 ` Andi Kleen 1 sibling, 3 replies; 35+ messages in thread From: Jeremy Fitzhardinge @ 2008-05-18 7:25 UTC (permalink / raw) To: Thomas Gleixner Cc: Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen Thomas Gleixner wrote: > Definitely, but we should do it at the Kconfig level which allows us > to have integer defines as well, so we end up with something like: > > static inline unsigned int get_nmi_count(int cpu) > { > return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); > } > Unfortunately that doesn't work because when CONFIG_X86_64 isn't defined it doesn't expand to 0. It would be nice if CONFIG_* expanded to 0/1, but we'd need to change all the #ifdef CONFIG_* to #if CONFIG_*... J ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 7:25 ` Jeremy Fitzhardinge @ 2008-05-18 7:38 ` Cyrill Gorcunov 2008-05-18 8:33 ` Jeremy Fitzhardinge 2008-05-18 9:09 ` Thomas Gleixner 2008-05-18 18:08 ` Adrian Bunk 2 siblings, 1 reply; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 7:38 UTC (permalink / raw) To: Jeremy Fitzhardinge Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen [Jeremy Fitzhardinge - Sun, May 18, 2008 at 08:25:38AM +0100] > Thomas Gleixner wrote: >> Definitely, but we should do it at the Kconfig level which allows us >> to have integer defines as well, so we end up with something like: >> >> static inline unsigned int get_nmi_count(int cpu) >> { >> return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); >> } >> > > Unfortunately that doesn't work because when CONFIG_X86_64 isn't defined it > doesn't expand to 0. It would be nice if CONFIG_* expanded to 0/1, but > we'd need to change all the #ifdef CONFIG_* to #if CONFIG_*... > > J > I think I would prefer Maciej's advice then but with capital letters (to easy distinguish them and point an attention) like #ifdef CONFG_X86_64 #define CPU_64 1 #else #define CPU_64 0 #endif - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 7:38 ` Cyrill Gorcunov @ 2008-05-18 8:33 ` Jeremy Fitzhardinge 2008-05-18 8:47 ` Cyrill Gorcunov 2008-05-18 9:13 ` Cyrill Gorcunov 0 siblings, 2 replies; 35+ messages in thread From: Jeremy Fitzhardinge @ 2008-05-18 8:33 UTC (permalink / raw) To: Cyrill Gorcunov Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen Cyrill Gorcunov wrote: > I think I would prefer Maciej's advice then but with capital > letters (to easy distinguish them and point an attention) like > > #ifdef CONFG_X86_64 > #define CPU_64 1 > #else > #define CPU_64 0 > #endif The other problem in this case is that cpu_pda() doesn't exit for 32-bit, so it probably won't compile anyway. But in general, I fully support using if (constant) over #ifdef (?: not so much). J ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 8:33 ` Jeremy Fitzhardinge @ 2008-05-18 8:47 ` Cyrill Gorcunov 2008-05-18 9:13 ` Cyrill Gorcunov 1 sibling, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 8:47 UTC (permalink / raw) To: Jeremy Fitzhardinge Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen [Jeremy Fitzhardinge - Sun, May 18, 2008 at 09:33:14AM +0100] > Cyrill Gorcunov wrote: >> I think I would prefer Maciej's advice then but with capital >> letters (to easy distinguish them and point an attention) like >> >> #ifdef CONFG_X86_64 >> #define CPU_64 1 >> #else >> #define CPU_64 0 >> #endif > > The other problem in this case is that cpu_pda() doesn't exit for 32-bit, > so it probably won't compile anyway. But in general, I fully support using > if (constant) over #ifdef (?: not so much). > > J > Since it's a compile-time known path I hope gcc just throw it out before checking for existance. But not tried yet ;) - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 8:33 ` Jeremy Fitzhardinge 2008-05-18 8:47 ` Cyrill Gorcunov @ 2008-05-18 9:13 ` Cyrill Gorcunov 1 sibling, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 9:13 UTC (permalink / raw) To: Jeremy Fitzhardinge Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen [Jeremy Fitzhardinge - Sun, May 18, 2008 at 09:33:14AM +0100] > Cyrill Gorcunov wrote: >> I think I would prefer Maciej's advice then but with capital >> letters (to easy distinguish them and point an attention) like >> >> #ifdef CONFG_X86_64 >> #define CPU_64 1 >> #else >> #define CPU_64 0 >> #endif > > The other problem in this case is that cpu_pda() doesn't exit for 32-bit, > so it probably won't compile anyway. But in general, I fully support using > if (constant) over #ifdef (?: not so much). > > J > yep, you're right - gcc doesn't throw it out but tries to evaluate in any case so this trick with () ? : ; would not work :( or we should define dummy types for this case... - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 7:25 ` Jeremy Fitzhardinge 2008-05-18 7:38 ` Cyrill Gorcunov @ 2008-05-18 9:09 ` Thomas Gleixner 2008-05-18 9:35 ` Cyrill Gorcunov 2008-05-18 18:08 ` Adrian Bunk 2 siblings, 1 reply; 35+ messages in thread From: Thomas Gleixner @ 2008-05-18 9:09 UTC (permalink / raw) To: Jeremy Fitzhardinge Cc: Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen On Sun, 18 May 2008, Jeremy Fitzhardinge wrote: > Thomas Gleixner wrote: > > Definitely, but we should do it at the Kconfig level which allows us > > to have integer defines as well, so we end up with something like: > > > > static inline unsigned int get_nmi_count(int cpu) > > { > > return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); > > } > > > > Unfortunately that doesn't work because when CONFIG_X86_64 isn't defined it > doesn't expand to 0. It would be nice if CONFIG_* expanded to 0/1, but we'd > need to change all the #ifdef CONFIG_* to #if CONFIG_*... You can have int type CONFIG_ which is always expanded. We have to add one of those though. Thanks, tglx ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 9:09 ` Thomas Gleixner @ 2008-05-18 9:35 ` Cyrill Gorcunov 0 siblings, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 9:35 UTC (permalink / raw) To: Thomas Gleixner Cc: Jeremy Fitzhardinge, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen [Thomas Gleixner - Sun, May 18, 2008 at 11:09:22AM +0200] | On Sun, 18 May 2008, Jeremy Fitzhardinge wrote: | | > Thomas Gleixner wrote: | > > Definitely, but we should do it at the Kconfig level which allows us | > > to have integer defines as well, so we end up with something like: | > > | > > static inline unsigned int get_nmi_count(int cpu) | > > { | > > return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); | > > } | > > | > | > Unfortunately that doesn't work because when CONFIG_X86_64 isn't defined it | > doesn't expand to 0. It would be nice if CONFIG_* expanded to 0/1, but we'd | > need to change all the #ifdef CONFIG_* to #if CONFIG_*... | | You can have int type CONFIG_ which is always expanded. We have to add | one of those though. | | Thanks, | tglx | Thomas, unfortunetly as I see we can't go by a simple way like that, these static funstions also hides the differen types and args list. So even we could leave it as it is now, or could define them as macroses. Anyway I'll try to find out how to handle this. - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 7:25 ` Jeremy Fitzhardinge 2008-05-18 7:38 ` Cyrill Gorcunov 2008-05-18 9:09 ` Thomas Gleixner @ 2008-05-18 18:08 ` Adrian Bunk 2008-05-18 18:13 ` Andi Kleen ` (2 more replies) 2 siblings, 3 replies; 35+ messages in thread From: Adrian Bunk @ 2008-05-18 18:08 UTC (permalink / raw) To: Jeremy Fitzhardinge Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen On Sun, May 18, 2008 at 08:25:38AM +0100, Jeremy Fitzhardinge wrote: > Thomas Gleixner wrote: >> Definitely, but we should do it at the Kconfig level which allows us >> to have integer defines as well, so we end up with something like: >> >> static inline unsigned int get_nmi_count(int cpu) >> { >> return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); >> } >> > > Unfortunately that doesn't work because when CONFIG_X86_64 isn't defined > it doesn't expand to 0. It would be nice if CONFIG_* expanded to 0/1, > but we'd need to change all the #ifdef CONFIG_* to #if CONFIG_*... Even more important: How do you want to handle kconfig variables set to "m"? Expand them to 0.5 ? ;-) > J cu Adrian -- "Is there not promise of rain?" Ling Tan asked suddenly out of the darkness. There had been need of rain for many days. "Only a promise," Lao Er said. Pearl S. Buck - Dragon Seed ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 18:08 ` Adrian Bunk @ 2008-05-18 18:13 ` Andi Kleen 2008-05-18 18:35 ` Jeremy Fitzhardinge 2008-05-18 18:33 ` Jeremy Fitzhardinge 2008-05-18 18:38 ` Maciej W. Rozycki 2 siblings, 1 reply; 35+ messages in thread From: Andi Kleen @ 2008-05-18 18:13 UTC (permalink / raw) To: Adrian Bunk Cc: Jeremy Fitzhardinge, Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg Adrian Bunk wrote: > On Sun, May 18, 2008 at 08:25:38AM +0100, Jeremy Fitzhardinge wrote: >> Thomas Gleixner wrote: >>> Definitely, but we should do it at the Kconfig level which allows us >>> to have integer defines as well, so we end up with something like: >>> >>> static inline unsigned int get_nmi_count(int cpu) >>> { >>> return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); >>> } >>> >> Unfortunately that doesn't work because when CONFIG_X86_64 isn't defined >> it doesn't expand to 0. It would be nice if CONFIG_* expanded to 0/1, >> but we'd need to change all the #ifdef CONFIG_* to #if CONFIG_*... > > Even more important: > How do you want to handle kconfig variables set to "m"? > > Expand them to 0.5 ? ;-) The whole idea was pretty bad. Ifdefs are not ugly because the syntax looks ugly, but because it's a semantically ugly construct with bad maintainability impact. Trying to put syntactical sugar around that is a doomed exercise. It will be still ugly, no matter what you do. -Andi ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 18:13 ` Andi Kleen @ 2008-05-18 18:35 ` Jeremy Fitzhardinge 2008-05-18 19:13 ` Andi Kleen 0 siblings, 1 reply; 35+ messages in thread From: Jeremy Fitzhardinge @ 2008-05-18 18:35 UTC (permalink / raw) To: Andi Kleen Cc: Adrian Bunk, Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg Andi Kleen wrote: > The whole idea was pretty bad. Ifdefs are not ugly because the syntax > looks ugly, but because it's a semantically ugly construct with bad > maintainability impact. > > Trying to put syntactical sugar around that is a doomed exercise. It > will be still ugly, no matter what you do. Not true. Using C rather than CPP to control the compilation of config options has the big win that all code paths are still visible to the compiler. In some cases that's not what you want, but it often is, and it would avoid some degree if inadvertent breakage of options. It can also be syntactically a lot more pleasant. J ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 18:35 ` Jeremy Fitzhardinge @ 2008-05-18 19:13 ` Andi Kleen 2008-05-19 14:27 ` Cyrill Gorcunov 0 siblings, 1 reply; 35+ messages in thread From: Andi Kleen @ 2008-05-18 19:13 UTC (permalink / raw) To: Jeremy Fitzhardinge Cc: Adrian Bunk, Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg Jeremy Fitzhardinge wrote: > Andi Kleen wrote: >> The whole idea was pretty bad. Ifdefs are not ugly because the syntax >> looks ugly, but because it's a semantically ugly construct with bad >> maintainability impact. >> >> Trying to put syntactical sugar around that is a doomed exercise. It >> will be still ugly, no matter what you do. > > Not true. Using C rather than CPP to control the compilation of config > options has the big win that all code paths are still visible to the > compiler. A small win. Still lots of other problems, including testing. In some cases that's not what you want, but it often is, and > it would avoid some degree if inadvertent breakage of options. It can > also be syntactically a lot more pleasant. Well it's still an unnecessary different code path and making it look nicer is just an excuse from properly cleaning it up. -Andi ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 19:13 ` Andi Kleen @ 2008-05-19 14:27 ` Cyrill Gorcunov 0 siblings, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-19 14:27 UTC (permalink / raw) To: Andi Kleen Cc: Jeremy Fitzhardinge, Adrian Bunk, Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg [Andi Kleen - Sun, May 18, 2008 at 09:13:04PM +0200] | Jeremy Fitzhardinge wrote: | > Andi Kleen wrote: | >> The whole idea was pretty bad. Ifdefs are not ugly because the syntax | >> looks ugly, but because it's a semantically ugly construct with bad | >> maintainability impact. | >> | >> Trying to put syntactical sugar around that is a doomed exercise. It | >> will be still ugly, no matter what you do. | > | > Not true. Using C rather than CPP to control the compilation of config | > options has the big win that all code paths are still visible to the | > compiler. | | A small win. Still lots of other problems, including testing. | | In some cases that's not what you want, but it often is, and | > it would avoid some degree if inadvertent breakage of options. It can | > also be syntactically a lot more pleasant. | | Well it's still an unnecessary different code path and making | it look nicer is just an excuse from properly cleaning it up. | | -Andi | ok, thanks to all reviewers for the look being taken. I will try to eliminate as much #ifdefs as possible, make them straight and understandable and will send the next patch as only get it done. Thanks to all! - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 18:08 ` Adrian Bunk 2008-05-18 18:13 ` Andi Kleen @ 2008-05-18 18:33 ` Jeremy Fitzhardinge 2008-05-18 19:29 ` Adrian Bunk 2008-05-18 18:38 ` Maciej W. Rozycki 2 siblings, 1 reply; 35+ messages in thread From: Jeremy Fitzhardinge @ 2008-05-18 18:33 UTC (permalink / raw) To: Adrian Bunk Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen Adrian Bunk wrote: > Even more important: > How do you want to handle kconfig variables set to "m"? > > Expand them to 0.5 ? ;-) > Doesn't much matter. We can't do "#ifdef CONFIG_FOO == m" anyway. J ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 18:33 ` Jeremy Fitzhardinge @ 2008-05-18 19:29 ` Adrian Bunk 2008-05-18 19:51 ` Jeremy Fitzhardinge 0 siblings, 1 reply; 35+ messages in thread From: Adrian Bunk @ 2008-05-18 19:29 UTC (permalink / raw) To: Jeremy Fitzhardinge Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen On Sun, May 18, 2008 at 07:33:47PM +0100, Jeremy Fitzhardinge wrote: > Adrian Bunk wrote: >> Even more important: >> How do you want to handle kconfig variables set to "m"? >> >> Expand them to 0.5 ? ;-) >> > > Doesn't much matter. We can't do "#ifdef CONFIG_FOO == m" anyway. Yes, we can: #ifdef CONFIG_FOO_MODULE > J cu Adrian -- "Is there not promise of rain?" Ling Tan asked suddenly out of the darkness. There had been need of rain for many days. "Only a promise," Lao Er said. Pearl S. Buck - Dragon Seed ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 19:29 ` Adrian Bunk @ 2008-05-18 19:51 ` Jeremy Fitzhardinge 0 siblings, 0 replies; 35+ messages in thread From: Jeremy Fitzhardinge @ 2008-05-18 19:51 UTC (permalink / raw) To: Adrian Bunk Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen Adrian Bunk wrote: > On Sun, May 18, 2008 at 07:33:47PM +0100, Jeremy Fitzhardinge wrote: > >> Adrian Bunk wrote: >> >>> Even more important: >>> How do you want to handle kconfig variables set to "m"? >>> >>> Expand them to 0.5 ? ;-) >>> >>> >> Doesn't much matter. We can't do "#ifdef CONFIG_FOO == m" anyway. >> > > Yes, we can: > > #ifdef CONFIG_FOO_MODULE > Well then, that's your answer. J ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 18:08 ` Adrian Bunk 2008-05-18 18:13 ` Andi Kleen 2008-05-18 18:33 ` Jeremy Fitzhardinge @ 2008-05-18 18:38 ` Maciej W. Rozycki 2008-05-18 20:40 ` Adrian Bunk 2 siblings, 1 reply; 35+ messages in thread From: Maciej W. Rozycki @ 2008-05-18 18:38 UTC (permalink / raw) To: Adrian Bunk Cc: Jeremy Fitzhardinge, Thomas Gleixner, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen On Sun, 18 May 2008, Adrian Bunk wrote: > Even more important: > How do you want to handle kconfig variables set to "m"? > > Expand them to 0.5 ? ;-) :-) Modules use a separate set of variables with _MODULE appended to their names. Thus for an option FOO, you'll get: #undef CONFIG_FOO #undef CONFIG_FOO_MODULE if it's set to "n", #define CONFIG_FOO 1 #undef CONFIG_FOO_MODULE if it's set to "y", and #undef CONFIG_FOO #define CONFIG_FOO_MODULE 1 if it's set to "m". Individual tests may check these both as they find appropriate. Maciej ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 18:38 ` Maciej W. Rozycki @ 2008-05-18 20:40 ` Adrian Bunk 0 siblings, 0 replies; 35+ messages in thread From: Adrian Bunk @ 2008-05-18 20:40 UTC (permalink / raw) To: Maciej W. Rozycki Cc: Jeremy Fitzhardinge, Thomas Gleixner, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg, Andi Kleen On Sun, May 18, 2008 at 07:38:41PM +0100, Maciej W. Rozycki wrote: > On Sun, 18 May 2008, Adrian Bunk wrote: > > > Even more important: > > How do you want to handle kconfig variables set to "m"? > > > > Expand them to 0.5 ? ;-) > > :-) > > Modules use a separate set of variables with _MODULE appended to their > names. Thus for an option FOO, you'll get: > > #undef CONFIG_FOO > #undef CONFIG_FOO_MODULE > > if it's set to "n", > > #define CONFIG_FOO 1 > #undef CONFIG_FOO_MODULE > > if it's set to "y", and > > #undef CONFIG_FOO > #define CONFIG_FOO_MODULE 1 > > if it's set to "m". Individual tests may check these both as they find > appropriate. I do know that. But your suggestion was: static inline unsigned int get_nmi_count(int cpu) { return cpu_x86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); } And Jeremy said "It would be nice if CONFIG_* expanded to 0/1". My point was simply that like most of the fun^Wproblems we already have in kconfig also here the tristate logic we need would make stuff more tricky. > Maciej cu Adrian -- "Is there not promise of rain?" Ling Tan asked suddenly out of the darkness. There had been need of rain for many days. "Only a promise," Lao Er said. Pearl S. Buck - Dragon Seed ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 21:40 ` Thomas Gleixner 2008-05-18 7:25 ` Jeremy Fitzhardinge @ 2008-05-18 10:15 ` Andi Kleen 2008-05-18 10:20 ` Cyrill Gorcunov 1 sibling, 1 reply; 35+ messages in thread From: Andi Kleen @ 2008-05-18 10:15 UTC (permalink / raw) To: Thomas Gleixner Cc: Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg > Definitely, but we should do it at the Kconfig level which allows us > to have integer defines as well, so we end up with something like: > > static inline unsigned int get_nmi_count(int cpu) > { > return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); > } #ifdef CONFIG_X86_64 would evaluate true even with CONFIG_X86_64 == 0 -Andi ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 10:15 ` Andi Kleen @ 2008-05-18 10:20 ` Cyrill Gorcunov 2008-05-18 10:25 ` Andi Kleen 0 siblings, 1 reply; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 10:20 UTC (permalink / raw) To: Andi Kleen Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg [Andi Kleen - Sun, May 18, 2008 at 12:15:32PM +0200] | | > Definitely, but we should do it at the Kconfig level which allows us | > to have integer defines as well, so we end up with something like: | > | > static inline unsigned int get_nmi_count(int cpu) | > { | > return CONFIG_X86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); | > } | | #ifdef CONFIG_X86_64 would evaluate true even with CONFIG_X86_64 == 0 | | -Andi | yes, but what to do with absence of __nmi_count on 32bit and die_nmi uses different number of args? gcc follows both pathes anyway trying to evaluate where I prefer it would not... I mean I've got errors on compiling procedue 'cause of different number of args for die_nmi used in 32bit mode. That is why I've asked Thomas if it possible to add "panic" boot option for 32bit mode and make it familiar with 64bit mode and merge them eventually. - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 10:20 ` Cyrill Gorcunov @ 2008-05-18 10:25 ` Andi Kleen 2008-05-18 10:29 ` Cyrill Gorcunov 0 siblings, 1 reply; 35+ messages in thread From: Andi Kleen @ 2008-05-18 10:25 UTC (permalink / raw) To: Cyrill Gorcunov Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg Cyrill Gorcunov wrote: > yes, but what to do with absence of __nmi_count on 32bit and die_nmi > uses different number of args? gcc follows both pathes anyway trying > to evaluate where I prefer it would not... I mean I've got errors > on compiling procedue 'cause of different number of args for die_nmi > used in 32bit mode. That is why I've asked Thomas if it possible to > add "panic" boot option for 32bit mode and make it familiar with 64bit > mode and merge them eventually. Sorry just pointed out why the Kconfig idea doesn't work, nothing more. If you want to avoid ifdefs then you have to unify the functionality first. Putting syntactical sugar on ifdefs doesn't make sense. I haven't kept track of the exact state of the code, but if the per cpu data macros are finally as efficient as the PDA you could move the nmi_count to per_cpu in both for once. -Andi ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 10:25 ` Andi Kleen @ 2008-05-18 10:29 ` Cyrill Gorcunov 2008-05-18 12:07 ` Tom Spink 0 siblings, 1 reply; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 10:29 UTC (permalink / raw) To: Andi Kleen Cc: Thomas Gleixner, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg [Andi Kleen - Sun, May 18, 2008 at 12:25:42PM +0200] | Cyrill Gorcunov wrote: | | > yes, but what to do with absence of __nmi_count on 32bit and die_nmi | > uses different number of args? gcc follows both pathes anyway trying | > to evaluate where I prefer it would not... I mean I've got errors | > on compiling procedue 'cause of different number of args for die_nmi | > used in 32bit mode. That is why I've asked Thomas if it possible to | > add "panic" boot option for 32bit mode and make it familiar with 64bit | > mode and merge them eventually. | | Sorry just pointed out why the Kconfig idea doesn't work, nothing more. | | If you want to avoid ifdefs then you have to unify the functionality | first. Putting syntactical sugar on ifdefs doesn't make sense. | | I haven't kept track of the exact state of the code, but if the per cpu | data macros are finally as efficient as the PDA you could move the | nmi_count to per_cpu in both for once. | | -Andi | ok, thanks - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 10:29 ` Cyrill Gorcunov @ 2008-05-18 12:07 ` Tom Spink 2008-05-18 12:10 ` Cyrill Gorcunov 0 siblings, 1 reply; 35+ messages in thread From: Tom Spink @ 2008-05-18 12:07 UTC (permalink / raw) To: Cyrill Gorcunov Cc: Andi Kleen, Thomas Gleixner, Maciej W. Rozycki, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg 2008/5/18 Cyrill Gorcunov <gorcunov@gmail.com>: > [Andi Kleen - Sun, May 18, 2008 at 12:25:42PM +0200] > | Cyrill Gorcunov wrote: > | > | > yes, but what to do with absence of __nmi_count on 32bit and die_nmi > | > uses different number of args? gcc follows both pathes anyway trying > | > to evaluate where I prefer it would not... I mean I've got errors > | > on compiling procedue 'cause of different number of args for die_nmi > | > used in 32bit mode. That is why I've asked Thomas if it possible to > | > add "panic" boot option for 32bit mode and make it familiar with 64bit > | > mode and merge them eventually. > | > | Sorry just pointed out why the Kconfig idea doesn't work, nothing more. > | > | If you want to avoid ifdefs then you have to unify the functionality > | first. Putting syntactical sugar on ifdefs doesn't make sense. > | > | I haven't kept track of the exact state of the code, but if the per cpu > | data macros are finally as efficient as the PDA you could move the > | nmi_count to per_cpu in both for once. > | > | -Andi > | > > ok, thanks > > - Cyrill - > It looks, though, that the unification of traps_{32,64}.c might help eliminate some of these conditionals. After a quick glance, at the traps source, certainly the __die_nmi helper might even be eliminated. -- Regards, Tom Spink ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-18 12:07 ` Tom Spink @ 2008-05-18 12:10 ` Cyrill Gorcunov 0 siblings, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 12:10 UTC (permalink / raw) To: Tom Spink Cc: Andi Kleen, Thomas Gleixner, Maciej W. Rozycki, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Sam Ravnborg [Tom Spink - Sun, May 18, 2008 at 01:07:37PM +0100] | 2008/5/18 Cyrill Gorcunov <gorcunov@gmail.com>: | > [Andi Kleen - Sun, May 18, 2008 at 12:25:42PM +0200] | > | Cyrill Gorcunov wrote: | > | | > | > yes, but what to do with absence of __nmi_count on 32bit and die_nmi | > | > uses different number of args? gcc follows both pathes anyway trying | > | > to evaluate where I prefer it would not... I mean I've got errors | > | > on compiling procedue 'cause of different number of args for die_nmi | > | > used in 32bit mode. That is why I've asked Thomas if it possible to | > | > add "panic" boot option for 32bit mode and make it familiar with 64bit | > | > mode and merge them eventually. | > | | > | Sorry just pointed out why the Kconfig idea doesn't work, nothing more. | > | | > | If you want to avoid ifdefs then you have to unify the functionality | > | first. Putting syntactical sugar on ifdefs doesn't make sense. | > | | > | I haven't kept track of the exact state of the code, but if the per cpu | > | data macros are finally as efficient as the PDA you could move the | > | nmi_count to per_cpu in both for once. | > | | > | -Andi | > | | > | > ok, thanks | > | > - Cyrill - | > | | It looks, though, that the unification of traps_{32,64}.c might help | eliminate some of these conditionals. After a quick glance, at the | traps source, certainly the __die_nmi helper might even be eliminated. | | -- | Regards, | Tom Spink | yes, I've already unified them so I use single form for both cases. I'll send updated version on the week probably. - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 20:52 ` Maciej W. Rozycki 2008-05-17 21:40 ` Thomas Gleixner @ 2008-05-17 21:48 ` Mikael Pettersson 2008-05-17 22:34 ` Thomas Gleixner 1 sibling, 1 reply; 35+ messages in thread From: Mikael Pettersson @ 2008-05-17 21:48 UTC (permalink / raw) To: Maciej W. Rozycki Cc: Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, Thomas Gleixner, LKML, Jiri Slaby, Andi Kleen Maciej W. Rozycki writes: > On Sat, 17 May 2008, Tom Spink wrote: > > > 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 > > } > > > > I know it introduces a lot of these conditionals, but at least there > > is one place to look for the get_nmi_count function, instead of > > searching for all variants of the function. > > Well, I suppose some header should provide a definition like: > > #ifdef CONFIG_X86_64 > #define cpu_x86_64 1 > #else > #define cpu_x86_64 0 > #endif > > and the you can remove the horrible #ifdef clutter and make the quoted > function look like: > > static inline unsigned int get_nmi_count(int cpu) > { > return cpu_x86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); > } > > Much better -- isn't it? IMO, no, the #ifdef is preferable. Why? Because the #ifdef is a very visible signal to the platform people that there are (in this case) subarch differences that force "clients" to behave differently on different subarchs. By removing the #ifdef you're IMO making it less likely for the platform people to take notice and work towards eliminating those differences. ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 21:48 ` Mikael Pettersson @ 2008-05-17 22:34 ` Thomas Gleixner 2008-05-18 6:24 ` Cyrill Gorcunov ` (3 more replies) 0 siblings, 4 replies; 35+ messages in thread From: Thomas Gleixner @ 2008-05-17 22:34 UTC (permalink / raw) To: Mikael Pettersson Cc: Maciej W. Rozycki, Tom Spink, Cyrill Gorcunov, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Andi Kleen On Sat, 17 May 2008, Mikael Pettersson wrote: > Maciej W. Rozycki writes: > > On Sat, 17 May 2008, Tom Spink wrote: > > > > > 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 > > > } > > > > > > I know it introduces a lot of these conditionals, but at least there > > > is one place to look for the get_nmi_count function, instead of > > > searching for all variants of the function. > > > > Well, I suppose some header should provide a definition like: > > > > #ifdef CONFIG_X86_64 > > #define cpu_x86_64 1 > > #else > > #define cpu_x86_64 0 > > #endif > > > > and the you can remove the horrible #ifdef clutter and make the quoted > > function look like: > > > > static inline unsigned int get_nmi_count(int cpu) > > { > > return cpu_x86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); > > } > > > > Much better -- isn't it? > > IMO, no, the #ifdef is preferable. > > Why? Because the #ifdef is a very visible signal to the platform > people that there are (in this case) subarch differences that force > "clients" to behave differently on different subarchs. By removing > the #ifdef you're IMO making it less likely for the platform people > to take notice and work towards eliminating those differences. The #ifdef is a poor choice. Maciej is damned right, that the single function with a clear distinction of the return value is better in terms of readability and maintenance. As I said before, We can make this more visible with an uppercase CONFIG_WHATEVER instaed of the innocent cpu_x86_64 one, but both solutions are better than #ifdefs and provide simple grepable patterns. The awareness of those differences does not depend at all on an #ifdef. Developers who are aware of the platform differences prefer a readable not ifdef poluted code base. People who need to be poked into the difference via an #ifdef are probably not those who can actually clean it up. Thanks, tglx ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 22:34 ` Thomas Gleixner @ 2008-05-18 6:24 ` Cyrill Gorcunov 2008-05-18 10:04 ` Cyrill Gorcunov ` (2 subsequent siblings) 3 siblings, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 6:24 UTC (permalink / raw) To: Thomas Gleixner Cc: Mikael Pettersson, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Andi Kleen [Thomas Gleixner - Sun, May 18, 2008 at 12:34:15AM +0200] | On Sat, 17 May 2008, Mikael Pettersson wrote: | > Maciej W. Rozycki writes: | > > On Sat, 17 May 2008, Tom Spink wrote: | > > | > > > 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 | > > > } | > > > | > > > I know it introduces a lot of these conditionals, but at least there | > > > is one place to look for the get_nmi_count function, instead of | > > > searching for all variants of the function. | > > | > > Well, I suppose some header should provide a definition like: | > > | > > #ifdef CONFIG_X86_64 | > > #define cpu_x86_64 1 | > > #else | > > #define cpu_x86_64 0 | > > #endif | > > | > > and the you can remove the horrible #ifdef clutter and make the quoted | > > function look like: | > > | > > static inline unsigned int get_nmi_count(int cpu) | > > { | > > return cpu_x86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); | > > } | > > | > > Much better -- isn't it? | > | > IMO, no, the #ifdef is preferable. | > | > Why? Because the #ifdef is a very visible signal to the platform | > people that there are (in this case) subarch differences that force | > "clients" to behave differently on different subarchs. By removing | > the #ifdef you're IMO making it less likely for the platform people | > to take notice and work towards eliminating those differences. | | The #ifdef is a poor choice. Maciej is damned right, that the single | function with a clear distinction of the return value is better in | terms of readability and maintenance. | | As I said before, We can make this more visible with an uppercase | CONFIG_WHATEVER instaed of the innocent cpu_x86_64 one, but both | solutions are better than #ifdefs and provide simple grepable | patterns. | | The awareness of those differences does not depend at all on an | #ifdef. Developers who are aware of the platform differences prefer a | readable not ifdef poluted code base. People who need to be poked into | the difference via an #ifdef are probably not those who can actually | clean it up. | | Thanks, | | tglx | Thanks to all for catching this nit. Actually I was using single function definition with #ifdef inside at first attempt. But the result was a too ugly imho, so I switched in the definition form you see now. If single defs is preferrable - no problem, will change it. Anyway, I found there are some additional patches in Ingo's tip tree so I have to remake my patch completely. So, as only I've got it done - will send to LKML again for your justice ;) - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 22:34 ` Thomas Gleixner 2008-05-18 6:24 ` Cyrill Gorcunov @ 2008-05-18 10:04 ` Cyrill Gorcunov 2008-05-18 10:09 ` Cyrill Gorcunov 2008-05-19 18:07 ` Cyrill Gorcunov 3 siblings, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 10:04 UTC (permalink / raw) To: Thomas Gleixner Cc: Mikael Pettersson, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Andi Kleen [Thomas Gleixner - Sun, May 18, 2008 at 12:34:15AM +0200] | On Sat, 17 May 2008, Mikael Pettersson wrote: | > Maciej W. Rozycki writes: | > > On Sat, 17 May 2008, Tom Spink wrote: | > > | > > > 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 | > > > } | > > > | > > > I know it introduces a lot of these conditionals, but at least there | > > > is one place to look for the get_nmi_count function, instead of | > > > searching for all variants of the function. | > > | > > Well, I suppose some header should provide a definition like: | > > | > > #ifdef CONFIG_X86_64 | > > #define cpu_x86_64 1 | > > #else | > > #define cpu_x86_64 0 | > > #endif | > > | > > and the you can remove the horrible #ifdef clutter and make the quoted | > > function look like: | > > | > > static inline unsigned int get_nmi_count(int cpu) | > > { | > > return cpu_x86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); | > > } | > > | > > Much better -- isn't it? | > | > IMO, no, the #ifdef is preferable. | > | > Why? Because the #ifdef is a very visible signal to the platform | > people that there are (in this case) subarch differences that force | > "clients" to behave differently on different subarchs. By removing | > the #ifdef you're IMO making it less likely for the platform people | > to take notice and work towards eliminating those differences. | | The #ifdef is a poor choice. Maciej is damned right, that the single | function with a clear distinction of the return value is better in | terms of readability and maintenance. | | As I said before, We can make this more visible with an uppercase | CONFIG_WHATEVER instaed of the innocent cpu_x86_64 one, but both | solutions are better than #ifdefs and provide simple grepable | patterns. | | The awareness of those differences does not depend at all on an | #ifdef. Developers who are aware of the platform differences prefer a | readable not ifdef poluted code base. People who need to be poked into | the difference via an #ifdef are probably not those who can actually | clean it up. | | Thanks, | | tglx | Eventually these helpers could look like this, objections? --- 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 void __die_nmi(char *str, struct pt_regs *regs, int do_panic) { #ifdef CONFIG_X86_64 die_nmi(str, regs, do_panic); #else die_nmi(regs, str); #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; } --- they are pretty ugly anyway ;) - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 22:34 ` Thomas Gleixner 2008-05-18 6:24 ` Cyrill Gorcunov 2008-05-18 10:04 ` Cyrill Gorcunov @ 2008-05-18 10:09 ` Cyrill Gorcunov 2008-05-19 18:07 ` Cyrill Gorcunov 3 siblings, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-18 10:09 UTC (permalink / raw) To: Thomas Gleixner Cc: Mikael Pettersson, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Andi Kleen [Thomas Gleixner - Sun, May 18, 2008 at 12:34:15AM +0200] | On Sat, 17 May 2008, Mikael Pettersson wrote: | > Maciej W. Rozycki writes: | > > On Sat, 17 May 2008, Tom Spink wrote: | > > | > > > 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 | > > > } | > > > | > > > I know it introduces a lot of these conditionals, but at least there | > > > is one place to look for the get_nmi_count function, instead of | > > > searching for all variants of the function. | > > | > > Well, I suppose some header should provide a definition like: | > > | > > #ifdef CONFIG_X86_64 | > > #define cpu_x86_64 1 | > > #else | > > #define cpu_x86_64 0 | > > #endif | > > | > > and the you can remove the horrible #ifdef clutter and make the quoted | > > function look like: | > > | > > static inline unsigned int get_nmi_count(int cpu) | > > { | > > return cpu_x86_64 ? cpu_pda(cpu)->__nmi_count : nmi_count(cpu); | > > } | > > | > > Much better -- isn't it? | > | > IMO, no, the #ifdef is preferable. | > | > Why? Because the #ifdef is a very visible signal to the platform | > people that there are (in this case) subarch differences that force | > "clients" to behave differently on different subarchs. By removing | > the #ifdef you're IMO making it less likely for the platform people | > to take notice and work towards eliminating those differences. | | The #ifdef is a poor choice. Maciej is damned right, that the single | function with a clear distinction of the return value is better in | terms of readability and maintenance. | | As I said before, We can make this more visible with an uppercase | CONFIG_WHATEVER instaed of the innocent cpu_x86_64 one, but both | solutions are better than #ifdefs and provide simple grepable | patterns. | | The awareness of those differences does not depend at all on an | #ifdef. Developers who are aware of the platform differences prefer a | readable not ifdef poluted code base. People who need to be poked into | the difference via an #ifdef are probably not those who can actually | clean it up. | | Thanks, | | tglx | Btw Thomas, is there any reason why can't we add do_panic in die_nmi on 32bit? And that would help us to add "panic" boot option on 32bits as well. - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-17 22:34 ` Thomas Gleixner ` (2 preceding siblings ...) 2008-05-18 10:09 ` Cyrill Gorcunov @ 2008-05-19 18:07 ` Cyrill Gorcunov 2008-05-19 18:41 ` Cyrill Gorcunov 3 siblings, 1 reply; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-19 18:07 UTC (permalink / raw) To: Thomas Gleixner Cc: Mikael Pettersson, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Andi Kleen Btw, if someone is still watching this thread - I've found a bit strange behaviour of nmi on 32bit platform. Look, we have nmi_watchdog = NMI_DEFAULT by default which is the alias to NMI_DISABLED. Then lets assume that user put some option on command line, say for example he passes something like that nmi_watchdog=2 which set it to nmi_watchdog = NMI_LOCAL_APIC with only that option passed we have sysfs entry created but I can't figure out why in proc_nmi_enabled() we have this code if (nmi_watchdog == NMI_DEFAULT) { if (lapic_watchdog_ok()) nmi_watchdog = NMI_LOCAL_APIC; else nmi_watchdog = NMI_IO_APIC; } it seems it just _dont need_ at and could be safetly removed. Did I miss something? - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-19 18:07 ` Cyrill Gorcunov @ 2008-05-19 18:41 ` Cyrill Gorcunov 2008-05-21 7:41 ` Cyrill Gorcunov 0 siblings, 1 reply; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-19 18:41 UTC (permalink / raw) To: Thomas Gleixner, Mikael Pettersson, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Andi Kleen [Cyrill Gorcunov - Mon, May 19, 2008 at 10:07:53PM +0400] | Btw, if someone is still watching this thread - I've found a bit strange | behaviour of nmi on 32bit platform. Look, we have | | nmi_watchdog = NMI_DEFAULT | | by default which is the alias to NMI_DISABLED. Then lets assume that | user put some option on command line, say for example he passes something | like that | | nmi_watchdog=2 | | which set it to | | nmi_watchdog = NMI_LOCAL_APIC | | with only that option passed we have sysfs entry created | but I can't figure out why in proc_nmi_enabled() we have this | code | | if (nmi_watchdog == NMI_DEFAULT) { | if (lapic_watchdog_ok()) | nmi_watchdog = NMI_LOCAL_APIC; | else | nmi_watchdog = NMI_IO_APIC; | } | | it seems it just _dont need_ at and could be safetly removed. | Did I miss something? | | - Cyrill - Ah, I found... sorry for noise, I'm shutting up ;) - Cyrill - ^ permalink raw reply [flat|nested] 35+ messages in thread
* Re: [RFC] x86: merge nmi_32-64 to nmi.c 2008-05-19 18:41 ` Cyrill Gorcunov @ 2008-05-21 7:41 ` Cyrill Gorcunov 0 siblings, 0 replies; 35+ messages in thread From: Cyrill Gorcunov @ 2008-05-21 7:41 UTC (permalink / raw) To: Thomas Gleixner, Mikael Pettersson, Maciej W. Rozycki, Tom Spink, Ingo Molnar, H. Peter Anvin, LKML, Jiri Slaby, Andi Kleen [-- Attachment #1: Type: text/plain, Size: 173 bytes --] well, here is a second version of patch attached (i'm in office now so can't use my lovely mutt as usual). Please take a look. Any objections are quite welcome! - Cyrill - [-- Attachment #2: nmi-32-64-merge-v2.patch --] [-- Type: application/octet-stream, Size: 41058 bytes --] [RFC] x86: merge nmi 32/64 v2 Here is a second version of merging attempt. Most notable changes from previous version: - definition of panic_on_unrecovered_nmi moved to traps_64.c (as traps_32.c do) - 32bit die_nmi changed to be more familiar to 64bit so we can use single definition (and pass do_panic if needed) - setup_nmi_watchdog now accept "nmi_watchdog=panic" boot option even in 32bit mode - we use apic_write_around for both cpu modes since it will be expanded to apic_write in 64bit mode anyway but save us from additional #ifdef Subroutines get_nmi_count(), mce_in_progress(), get_irq_sum() can't eliminate #ifdef using since they do HIDE different types. To eliminate such differencies completely much more deep changes has to be brought which I think would not be good at once. So I don't see a simple way to use CONFIG_ vars with following ? : constructions to strip #ifdef out. Any _OBJECTIONS_ are welcome! Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com> --- Index: arch/x86/kernel/nmi.c ===================================================================== --- /dev/null Wed May 21 10:11:28 2008 +++ b/arch/x86/kernel/nmi.c Wed May 21 10:06:28 2008 @@ -0,0 +1,551 @@ +/* + * NMI watchdog support on APIC systems + * + * Started by Ingo Molnar <mingo@redhat.com> + * + * 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 <linux/nmi.h> +#include <linux/mm.h> +#include <linux/delay.h> +#include <linux/interrupt.h> +#include <linux/module.h> +#include <linux/sysdev.h> +#include <linux/sysctl.h> +#include <linux/kprobes.h> +#include <linux/cpumask.h> +#include <linux/kdebug.h> +#include <linux/kernel_stat.h> + +#include <asm/smp.h> +#include <asm/nmi.h> +#include <asm/proto.h> +#include <asm/mce.h> +#include <asm/timer.h> + +#include <mach_traps.h> + +int unknown_nmi_panic; +int nmi_watchdog_enabled; + +static int panic_on_timeout; + +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 */ +unsigned int nmi_watchdog = NMI_DEFAULT; + +static DEFINE_PER_CPU(short, wd_enabled); +static unsigned int nmi_hz = HZ; +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_irq_sum(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 /* CONFIG_SMP */ + +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 for 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: + __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_irq_sum(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); + 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 do + */ + 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("NMI Watchdog detected LOCKUP\n", + 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_X86_64 +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(); +} +#endif /* CONFIG_X86_64 */ + +#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); + 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); + Index: arch/x86/kernel/traps_64.c ===================================================================== --- a/arch/x86/kernel/traps_64.c Mon May 19 01:36:42 2008 +++ b/arch/x86/kernel/traps_64.c Wed May 21 08:50:40 2008 @@ -76,6 +76,7 @@ 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 inline void conditional_sti(struct pt_regs *regs) Index: arch/x86/kernel/traps_32.c ===================================================================== --- a/arch/x86/kernel/traps_32.c Mon May 19 01:36:42 2008 +++ b/arch/x86/kernel/traps_32.c Wed May 21 09:27:42 2008 @@ -755,9 +755,9 @@ unknown_nmi_error(unsigned char reason, 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 * 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); Index: include/asm-x86/nmi.h ===================================================================== --- a/include/asm-x86/nmi.h Mon May 19 01:36:42 2008 +++ b/include/asm-x86/nmi.h Wed May 21 09:28:20 2008 @@ -38,12 +38,10 @@ static inline void unset_nmi_pm_callback #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 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; extern int unknown_nmi_panic; Index: arch/x86/kernel/Makefile ===================================================================== --- a/arch/x86/kernel/Makefile Mon May 19 01:36:42 2008 +++ b/arch/x86/kernel/Makefile Wed May 21 10:17:42 2008 @@ -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 Index: arch/x86/kernel/nmi_32.c ===================================================================== --- a/x86/kernel/nmi_32.c Mon May 19 01:36:42 2008 +++ /dev/null Wed May 21 10:12:30 2008 @@ -1,472 +0,0 @@ -/* - * NMI watchdog support on APIC systems - * - * Started by Ingo Molnar <mingo@redhat.com> - * - * 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 <linux/delay.h> -#include <linux/interrupt.h> -#include <linux/module.h> -#include <linux/nmi.h> -#include <linux/sysdev.h> -#include <linux/sysctl.h> -#include <linux/percpu.h> -#include <linux/kprobes.h> -#include <linux/cpumask.h> -#include <linux/kernel_stat.h> -#include <linux/kdebug.h> -#include <linux/slab.h> - -#include <asm/smp.h> -#include <asm/nmi.h> -#include <asm/timer.h> - -#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 */ - -unsigned int nmi_watchdog = NMI_DEFAULT; -static unsigned int nmi_hz = HZ; - -static DEFINE_PER_CPU(short, wd_enabled); - -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 - * 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] = nmi_count(cpu); - 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 - if (!per_cpu(wd_enabled, cpu)) - continue; - if (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)); - 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; - - 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(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 unsigned int - last_irq_sums [NR_CPUS], - alert_counter [NR_CPUS]; - -void touch_nmi_watchdog(void) -{ - if (nmi_watchdog > 0) { - unsigned cpu; - - /* - * Just reset the alert counters, (other CPUs might be - * spinning on locks we hold): - */ - for_each_present_cpu(cpu) { - if (alert_counter[cpu]) - alert_counter[cpu] = 0; - } - } - - /* - * Tickle the softlockup detector too: - */ - touch_softlockup_watchdog(); -} -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) -{ - - /* - * 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; - } - - 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); - } - - /* - * 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 the none of the timers isn't firing, this cpu isn't doing much */ - if (!touched && last_irq_sums[cpu] == 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) - /* - * 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; - } - /* 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(regs, buf); - 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 == NMI_DEFAULT) { - if (lapic_watchdog_ok()) - nmi_watchdog = NMI_LOCAL_APIC; - else - nmi_watchdog = NMI_IO_APIC; - } - - 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); Index: arch/x86/kernel/nmi_64.c ===================================================================== --- a/x86/kernel/nmi_64.c Mon May 19 01:36:42 2008 +++ /dev/null Wed May 21 10:12:30 2008 @@ -1,482 +0,0 @@ -/* - * NMI watchdog support on APIC systems - * - * Started by Ingo Molnar <mingo@redhat.com> - * - * 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 <linux/nmi.h> -#include <linux/mm.h> -#include <linux/delay.h> -#include <linux/interrupt.h> -#include <linux/module.h> -#include <linux/sysdev.h> -#include <linux/sysctl.h> -#include <linux/kprobes.h> -#include <linux/cpumask.h> -#include <linux/kdebug.h> - -#include <asm/smp.h> -#include <asm/nmi.h> -#include <asm/proto.h> -#include <asm/mce.h> - -#include <mach_traps.h> - -int unknown_nmi_panic; -int nmi_watchdog_enabled; -int panic_on_unrecovered_nmi; - -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); - -/* 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; -} - -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 - * 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) -{ - 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 (cpu = 0; cpu < NR_CPUS; cpu++) - prev_nmi_count[cpu] = cpu_pda(cpu)->__nmi_count; - 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) { - printk(KERN_WARNING "WARNING: CPU#%d: NMI " - "appears to be stuck (%d->%d)!\n", - cpu, - prev_nmi_count[cpu], - cpu_pda(cpu)->__nmi_count); - 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(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(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) -{ - 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 = read_pda(apic_timer_irqs) + read_pda(irq0_irqs); - 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); - } - -#ifdef CONFIG_X86_MCE - /* Could check oops_in_progress here too, but it's safer - not too */ - if (atomic_read(&mce_entry) > 0) - 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) { - /* - * 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 on CPU %d\n", 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; -} - -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) -{ - 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); ^ permalink raw reply [flat|nested] 35+ messages in thread
end of thread, other threads:[~2008-05-21 7:41 UTC | newest] Thread overview: 35+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2008-05-17 19:22 [RFC] x86: merge nmi_32-64 to nmi.c Cyrill Gorcunov 2008-05-17 20:28 ` Tom Spink 2008-05-17 20:52 ` Maciej W. Rozycki 2008-05-17 21:40 ` Thomas Gleixner 2008-05-18 7:25 ` Jeremy Fitzhardinge 2008-05-18 7:38 ` Cyrill Gorcunov 2008-05-18 8:33 ` Jeremy Fitzhardinge 2008-05-18 8:47 ` Cyrill Gorcunov 2008-05-18 9:13 ` Cyrill Gorcunov 2008-05-18 9:09 ` Thomas Gleixner 2008-05-18 9:35 ` Cyrill Gorcunov 2008-05-18 18:08 ` Adrian Bunk 2008-05-18 18:13 ` Andi Kleen 2008-05-18 18:35 ` Jeremy Fitzhardinge 2008-05-18 19:13 ` Andi Kleen 2008-05-19 14:27 ` Cyrill Gorcunov 2008-05-18 18:33 ` Jeremy Fitzhardinge 2008-05-18 19:29 ` Adrian Bunk 2008-05-18 19:51 ` Jeremy Fitzhardinge 2008-05-18 18:38 ` Maciej W. Rozycki 2008-05-18 20:40 ` Adrian Bunk 2008-05-18 10:15 ` Andi Kleen 2008-05-18 10:20 ` Cyrill Gorcunov 2008-05-18 10:25 ` Andi Kleen 2008-05-18 10:29 ` Cyrill Gorcunov 2008-05-18 12:07 ` Tom Spink 2008-05-18 12:10 ` Cyrill Gorcunov 2008-05-17 21:48 ` Mikael Pettersson 2008-05-17 22:34 ` Thomas Gleixner 2008-05-18 6:24 ` Cyrill Gorcunov 2008-05-18 10:04 ` Cyrill Gorcunov 2008-05-18 10:09 ` Cyrill Gorcunov 2008-05-19 18:07 ` Cyrill Gorcunov 2008-05-19 18:41 ` Cyrill Gorcunov 2008-05-21 7:41 ` Cyrill Gorcunov
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox