[git pull] timer fixes

From: Ingo Molnar
Date: Tue Sep 23 2008 - 15:41:30 EST

Linus,

Please pull the latest timers-fixes-for-linus git tree from:

git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git timers-fixes-for-linus

A category of late fixes that solves the problems around regression
http://bugzilla.kernel.org/show_bug.cgi?id=11611 . Tested by Rafael, and
tested through on a range of laptops we have access to.

Thanks,

Ingo

------------------>
Andreas Herrmann (1):
x86: c1e_idle: don't mark TSC unstable if CPU has invariant TSC

Ingo Molnar (1):
timers: fix build error in !oneshot case

Thomas Gleixner (6):
clockevents: prevent cpu online to interfere with nohz
x86: prevent stale state of c1e_mask across CPU offline/online
clockevents: prevent stale tick_next_period for onlining CPUs
clockevents: check broadcast device not tick device
clockevents: prevent mode mismatch on cpu online
x86: prevent C-states hang on AMD C1E enabled machines


arch/x86/kernel/process.c | 17 ++++++++++++-----
arch/x86/kernel/process_32.c | 1 +
arch/x86/kernel/process_64.c | 2 ++
include/asm-x86/acpi.h | 2 ++
include/asm-x86/cpufeature.h | 1 +
include/asm-x86/idle.h | 2 ++
kernel/time/tick-broadcast.c | 12 ++++++++++--
kernel/time/tick-common.c | 10 ++++++----
kernel/time/tick-internal.h | 7 +++++++
kernel/time/tick-sched.c | 11 +++++++----
10 files changed, 50 insertions(+), 15 deletions(-)

diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 7fc4d5b..876e918 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -246,6 +246,14 @@ static int __cpuinit check_c1e_idle(const struct cpuinfo_x86 *c)
return 1;
}

+static cpumask_t c1e_mask = CPU_MASK_NONE;
+static int c1e_detected;
+
+void c1e_remove_cpu(int cpu)
+{
+ cpu_clear(cpu, c1e_mask);
+}
+
/*
* C1E aware idle routine. We check for C1E active in the interrupt
* pending message MSR. If we detect C1E, then we handle it the same
@@ -253,9 +261,6 @@ static int __cpuinit check_c1e_idle(const struct cpuinfo_x86 *c)
*/
static void c1e_idle(void)
{
- static cpumask_t c1e_mask = CPU_MASK_NONE;
- static int c1e_detected;
-
if (need_resched())
return;

@@ -265,8 +270,10 @@ static void c1e_idle(void)
rdmsr(MSR_K8_INT_PENDING_MSG, lo, hi);
if (lo & K8_INTP_C1E_ACTIVE_MASK) {
c1e_detected = 1;
- mark_tsc_unstable("TSC halt in C1E");
- printk(KERN_INFO "System has C1E enabled\n");
+ if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
+ mark_tsc_unstable("TSC halt in AMD C1E");
+ printk(KERN_INFO "System has AMD C1E enabled\n");
+ set_cpu_cap(&boot_cpu_data, X86_FEATURE_AMDC1E);
}
}

diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index 3b7a1dd..4b3cfdf 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -88,6 +88,7 @@ static void cpu_exit_clear(void)
cpu_clear(cpu, cpu_callin_map);

numa_remove_cpu(cpu);
+ c1e_remove_cpu(cpu);
}

/* We don't actually take CPU down, just spin without interrupts. */
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 71553b6..e12e0e4 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -93,6 +93,8 @@ DECLARE_PER_CPU(int, cpu_state);
static inline void play_dead(void)
{
idle_task_exit();
+ c1e_remove_cpu(raw_smp_processor_id());
+
mb();
/* Ack it */
__get_cpu_var(cpu_state) = CPU_DEAD;
diff --git a/include/asm-x86/acpi.h b/include/asm-x86/acpi.h
index 635d764..35d1743 100644
--- a/include/asm-x86/acpi.h
+++ b/include/asm-x86/acpi.h
@@ -140,6 +140,8 @@ static inline unsigned int acpi_processor_cstate_check(unsigned int max_cstate)
boot_cpu_data.x86_model <= 0x05 &&
boot_cpu_data.x86_mask < 0x0A)
return 1;
+ else if (boot_cpu_has(X86_FEATURE_AMDC1E))
+ return 1;
else
return max_cstate;
}
diff --git a/include/asm-x86/cpufeature.h b/include/asm-x86/cpufeature.h
index 9489283..cfcfb0a 100644
--- a/include/asm-x86/cpufeature.h
+++ b/include/asm-x86/cpufeature.h
@@ -81,6 +81,7 @@
#define X86_FEATURE_LFENCE_RDTSC (3*32+18) /* Lfence synchronizes RDTSC */
#define X86_FEATURE_11AP (3*32+19) /* Bad local APIC aka 11AP */
#define X86_FEATURE_NOPL (3*32+20) /* The NOPL (0F 1F) instructions */
+#define X86_FEATURE_AMDC1E (3*32+21) /* AMD C1E detected */

/* Intel-defined CPU features, CPUID level 0x00000001 (ecx), word 4 */
#define X86_FEATURE_XMM3 (4*32+ 0) /* Streaming SIMD Extensions-3 */
diff --git a/include/asm-x86/idle.h b/include/asm-x86/idle.h
index d240e5b..cbb6491 100644
--- a/include/asm-x86/idle.h
+++ b/include/asm-x86/idle.h
@@ -10,4 +10,6 @@ void idle_notifier_register(struct notifier_block *n);
void enter_idle(void);
void exit_idle(void);

+void c1e_remove_cpu(int cpu);
+
#endif
diff --git a/kernel/time/tick-broadcast.c b/kernel/time/tick-broadcast.c
index f1f3eee..bd70345 100644
--- a/kernel/time/tick-broadcast.c
+++ b/kernel/time/tick-broadcast.c
@@ -235,7 +235,7 @@ static void tick_do_broadcast_on_off(void *why)
case CLOCK_EVT_NOTIFY_BROADCAST_FORCE:
if (!cpu_isset(cpu, tick_broadcast_mask)) {
cpu_set(cpu, tick_broadcast_mask);
- if (td->mode == TICKDEV_MODE_PERIODIC)
+ if (bc->mode == TICKDEV_MODE_PERIODIC)
clockevents_shutdown(dev);
}
if (*reason == CLOCK_EVT_NOTIFY_BROADCAST_FORCE)
@@ -245,7 +245,7 @@ static void tick_do_broadcast_on_off(void *why)
if (!tick_broadcast_force &&
cpu_isset(cpu, tick_broadcast_mask)) {
cpu_clear(cpu, tick_broadcast_mask);
- if (td->mode == TICKDEV_MODE_PERIODIC)
+ if (bc->mode == TICKDEV_MODE_PERIODIC)
tick_setup_periodic(dev, 0);
}
break;
@@ -575,4 +575,12 @@ void tick_shutdown_broadcast_oneshot(unsigned int *cpup)
spin_unlock_irqrestore(&tick_broadcast_lock, flags);
}

+/*
+ * Check, whether the broadcast device is in one shot mode
+ */
+int tick_broadcast_oneshot_active(void)
+{
+ return tick_broadcast_device.mode == TICKDEV_MODE_ONESHOT;
+}
+
#endif
diff --git a/kernel/time/tick-common.c b/kernel/time/tick-common.c
index 019315e..df12434 100644
--- a/kernel/time/tick-common.c
+++ b/kernel/time/tick-common.c
@@ -33,7 +33,7 @@ DEFINE_PER_CPU(struct tick_device, tick_cpu_device);
*/
ktime_t tick_next_period;
ktime_t tick_period;
-int tick_do_timer_cpu __read_mostly = -1;
+int tick_do_timer_cpu __read_mostly = TICK_DO_TIMER_BOOT;
DEFINE_SPINLOCK(tick_device_lock);

/*
@@ -109,7 +109,8 @@ void tick_setup_periodic(struct clock_event_device *dev, int broadcast)
if (!tick_device_is_functional(dev))
return;

- if (dev->features & CLOCK_EVT_FEAT_PERIODIC) {
+ if ((dev->features & CLOCK_EVT_FEAT_PERIODIC) &&
+ !tick_broadcast_oneshot_active()) {
clockevents_set_mode(dev, CLOCK_EVT_MODE_PERIODIC);
} else {
unsigned long seq;
@@ -148,7 +149,7 @@ static void tick_setup_device(struct tick_device *td,
* If no cpu took the do_timer update, assign it to
* this cpu:
*/
- if (tick_do_timer_cpu == -1) {
+ if (tick_do_timer_cpu == TICK_DO_TIMER_BOOT) {
tick_do_timer_cpu = cpu;
tick_next_period = ktime_get();
tick_period = ktime_set(0, NSEC_PER_SEC / HZ);
@@ -300,7 +301,8 @@ static void tick_shutdown(unsigned int *cpup)
if (*cpup == tick_do_timer_cpu) {
int cpu = first_cpu(cpu_online_map);

- tick_do_timer_cpu = (cpu != NR_CPUS) ? cpu : -1;
+ tick_do_timer_cpu = (cpu != NR_CPUS) ? cpu :
+ TICK_DO_TIMER_NONE;
}
spin_unlock_irqrestore(&tick_device_lock, flags);
}
diff --git a/kernel/time/tick-internal.h b/kernel/time/tick-internal.h
index 6e9db97..4692487 100644
--- a/kernel/time/tick-internal.h
+++ b/kernel/time/tick-internal.h
@@ -1,6 +1,10 @@
/*
* tick internal variable and functions used by low/high res code
*/
+
+#define TICK_DO_TIMER_NONE -1
+#define TICK_DO_TIMER_BOOT -2
+
DECLARE_PER_CPU(struct tick_device, tick_cpu_device);
extern spinlock_t tick_device_lock;
extern ktime_t tick_next_period;
@@ -31,6 +35,7 @@ extern void tick_broadcast_oneshot_control(unsigned long reason);
extern void tick_broadcast_switch_to_oneshot(void);
extern void tick_shutdown_broadcast_oneshot(unsigned int *cpup);
extern int tick_resume_broadcast_oneshot(struct clock_event_device *bc);
+extern int tick_broadcast_oneshot_active(void);
# else /* BROADCAST */
static inline void tick_broadcast_setup_oneshot(struct clock_event_device *bc)
{
@@ -39,6 +44,7 @@ static inline void tick_broadcast_setup_oneshot(struct clock_event_device *bc)
static inline void tick_broadcast_oneshot_control(unsigned long reason) { }
static inline void tick_broadcast_switch_to_oneshot(void) { }
static inline void tick_shutdown_broadcast_oneshot(unsigned int *cpup) { }
+static inline int tick_broadcast_oneshot_active(void) { return 0; }
# endif /* !BROADCAST */

#else /* !ONESHOT */
@@ -68,6 +74,7 @@ static inline int tick_resume_broadcast_oneshot(struct clock_event_device *bc)
{
return 0;
}
+static inline int tick_broadcast_oneshot_active(void) { return 0; }
#endif /* !TICK_ONESHOT */

/*
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index a87b046..39019b3 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -75,6 +75,9 @@ static void tick_do_update_jiffies64(ktime_t now)
incr * ticks);
}
do_timer(++ticks);
+
+ /* Keep the tick_next_period variable up to date */
+ tick_next_period = ktime_add(last_jiffies_update, tick_period);
}
write_sequnlock(&xtime_lock);
}
@@ -221,7 +224,7 @@ void tick_nohz_stop_sched_tick(int inidle)
*/
if (unlikely(!cpu_online(cpu))) {
if (cpu == tick_do_timer_cpu)
- tick_do_timer_cpu = -1;
+ tick_do_timer_cpu = TICK_DO_TIMER_NONE;
}

if (unlikely(ts->nohz_mode == NOHZ_MODE_INACTIVE))
@@ -303,7 +306,7 @@ void tick_nohz_stop_sched_tick(int inidle)
* invoked.
*/
if (cpu == tick_do_timer_cpu)
- tick_do_timer_cpu = -1;
+ tick_do_timer_cpu = TICK_DO_TIMER_NONE;

ts->idle_sleeps++;

@@ -468,7 +471,7 @@ static void tick_nohz_handler(struct clock_event_device *dev)
* this duty, then the jiffies update is still serialized by
* xtime_lock.
*/
- if (unlikely(tick_do_timer_cpu == -1))
+ if (unlikely(tick_do_timer_cpu == TICK_DO_TIMER_NONE))
tick_do_timer_cpu = cpu;

/* Check, if the jiffies need an update */
@@ -570,7 +573,7 @@ static enum hrtimer_restart tick_sched_timer(struct hrtimer *timer)
* this duty, then the jiffies update is still serialized by
* xtime_lock.
*/
- if (unlikely(tick_do_timer_cpu == -1))
+ if (unlikely(tick_do_timer_cpu == TICK_DO_TIMER_NONE))
tick_do_timer_cpu = cpu;
#endif

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