linux/arch/x86/kernel/cpu/perf_counter.c

734 lines
18 KiB
C
Raw Normal View History

/*
* Performance counter x86 architecture code
*
* Copyright(C) 2008 Thomas Gleixner <tglx@linutronix.de>
* Copyright(C) 2008 Red Hat, Inc., Ingo Molnar
*
* For licencing details see kernel-base/COPYING
*/
#include <linux/perf_counter.h>
#include <linux/capability.h>
#include <linux/notifier.h>
#include <linux/hardirq.h>
#include <linux/kprobes.h>
#include <linux/module.h>
#include <linux/kdebug.h>
#include <linux/sched.h>
#include <asm/perf_counter.h>
#include <asm/apic.h>
static bool perf_counters_initialized __read_mostly;
/*
* Number of (generic) HW counters:
*/
static int nr_counters_generic __read_mostly;
static u64 perf_counter_mask __read_mostly;
static u64 counter_value_mask __read_mostly;
static int nr_counters_fixed __read_mostly;
struct cpu_hw_counters {
struct perf_counter *counters[X86_PMC_IDX_MAX];
unsigned long used[BITS_TO_LONGS(X86_PMC_IDX_MAX)];
unsigned long interrupts;
u64 global_enable;
};
/*
* Intel PerfMon v3. Used on Core2 and later.
*/
static DEFINE_PER_CPU(struct cpu_hw_counters, cpu_hw_counters);
static const int intel_perfmon_event_map[] =
{
[PERF_COUNT_CPU_CYCLES] = 0x003c,
[PERF_COUNT_INSTRUCTIONS] = 0x00c0,
[PERF_COUNT_CACHE_REFERENCES] = 0x4f2e,
[PERF_COUNT_CACHE_MISSES] = 0x412e,
[PERF_COUNT_BRANCH_INSTRUCTIONS] = 0x00c4,
[PERF_COUNT_BRANCH_MISSES] = 0x00c5,
[PERF_COUNT_BUS_CYCLES] = 0x013c,
};
static const int max_intel_perfmon_events = ARRAY_SIZE(intel_perfmon_event_map);
/*
* Propagate counter elapsed time into the generic counter.
* Can only be executed on the CPU where the counter is active.
* Returns the delta events processed.
*/
static void
x86_perf_counter_update(struct perf_counter *counter,
struct hw_perf_counter *hwc, int idx)
{
u64 prev_raw_count, new_raw_count, delta;
/*
* Careful: an NMI might modify the previous counter value.
*
* Our tactic to handle this is to first atomically read and
* exchange a new raw count - then add that new-prev delta
* count to the generic counter atomically:
*/
again:
prev_raw_count = atomic64_read(&hwc->prev_count);
rdmsrl(hwc->counter_base + idx, new_raw_count);
if (atomic64_cmpxchg(&hwc->prev_count, prev_raw_count,
new_raw_count) != prev_raw_count)
goto again;
/*
* Now we have the new raw value and have updated the prev
* timestamp already. We can now calculate the elapsed delta
* (counter-)time and add that to the generic counter.
*
* Careful, not all hw sign-extends above the physical width
* of the count, so we do that by clipping the delta to 32 bits:
*/
delta = (u64)(u32)((s32)new_raw_count - (s32)prev_raw_count);
atomic64_add(delta, &counter->count);
atomic64_sub(delta, &hwc->period_left);
}
/*
* Setup the hardware configuration for a given hw_event_type
*/
static int __hw_perf_counter_init(struct perf_counter *counter)
{
struct perf_counter_hw_event *hw_event = &counter->hw_event;
struct hw_perf_counter *hwc = &counter->hw;
if (unlikely(!perf_counters_initialized))
return -EINVAL;
/*
perf_counters: allow users to count user, kernel and/or hypervisor events Impact: new perf_counter feature This extends the perf_counter_hw_event struct with bits that specify that events in user, kernel and/or hypervisor mode should not be counted (i.e. should be excluded), and adds code to program the PMU mode selection bits accordingly on x86 and powerpc. For software counters, we don't currently have the infrastructure to distinguish which mode an event occurs in, so we currently fail the counter initialization if the setting of the hw_event.exclude_* bits would require us to distinguish. Context switches and CPU migrations are currently considered to occur in kernel mode. On x86, this changes the previous policy that only root can count kernel events. Now non-root users can count kernel events or exclude them. Non-root users still can't use NMI events, though. On x86 we don't appear to have any way to control whether hypervisor events are counted or not, so hw_event.exclude_hv is ignored. On powerpc, the selection of whether to count events in user, kernel and/or hypervisor mode is PMU-wide, not per-counter, so this adds a check that the hw_event.exclude_* settings are the same as other events on the PMU. Counters being added to a group have to have the same settings as the other hardware counters in the group. Counters and groups can only be enabled in hw_perf_group_sched_in or power_perf_enable if they have the same settings as any other counters already on the PMU. If we are not running on a hypervisor, the exclude_hv setting is ignored (by forcing it to 0) since we can't ever get any hypervisor events. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-11 03:35:35 +00:00
* Generate PMC IRQs:
* (keep 'enabled' bit clear for now)
*/
perf_counters: allow users to count user, kernel and/or hypervisor events Impact: new perf_counter feature This extends the perf_counter_hw_event struct with bits that specify that events in user, kernel and/or hypervisor mode should not be counted (i.e. should be excluded), and adds code to program the PMU mode selection bits accordingly on x86 and powerpc. For software counters, we don't currently have the infrastructure to distinguish which mode an event occurs in, so we currently fail the counter initialization if the setting of the hw_event.exclude_* bits would require us to distinguish. Context switches and CPU migrations are currently considered to occur in kernel mode. On x86, this changes the previous policy that only root can count kernel events. Now non-root users can count kernel events or exclude them. Non-root users still can't use NMI events, though. On x86 we don't appear to have any way to control whether hypervisor events are counted or not, so hw_event.exclude_hv is ignored. On powerpc, the selection of whether to count events in user, kernel and/or hypervisor mode is PMU-wide, not per-counter, so this adds a check that the hw_event.exclude_* settings are the same as other events on the PMU. Counters being added to a group have to have the same settings as the other hardware counters in the group. Counters and groups can only be enabled in hw_perf_group_sched_in or power_perf_enable if they have the same settings as any other counters already on the PMU. If we are not running on a hypervisor, the exclude_hv setting is ignored (by forcing it to 0) since we can't ever get any hypervisor events. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-11 03:35:35 +00:00
hwc->config = ARCH_PERFMON_EVENTSEL_INT;
/*
perf_counters: allow users to count user, kernel and/or hypervisor events Impact: new perf_counter feature This extends the perf_counter_hw_event struct with bits that specify that events in user, kernel and/or hypervisor mode should not be counted (i.e. should be excluded), and adds code to program the PMU mode selection bits accordingly on x86 and powerpc. For software counters, we don't currently have the infrastructure to distinguish which mode an event occurs in, so we currently fail the counter initialization if the setting of the hw_event.exclude_* bits would require us to distinguish. Context switches and CPU migrations are currently considered to occur in kernel mode. On x86, this changes the previous policy that only root can count kernel events. Now non-root users can count kernel events or exclude them. Non-root users still can't use NMI events, though. On x86 we don't appear to have any way to control whether hypervisor events are counted or not, so hw_event.exclude_hv is ignored. On powerpc, the selection of whether to count events in user, kernel and/or hypervisor mode is PMU-wide, not per-counter, so this adds a check that the hw_event.exclude_* settings are the same as other events on the PMU. Counters being added to a group have to have the same settings as the other hardware counters in the group. Counters and groups can only be enabled in hw_perf_group_sched_in or power_perf_enable if they have the same settings as any other counters already on the PMU. If we are not running on a hypervisor, the exclude_hv setting is ignored (by forcing it to 0) since we can't ever get any hypervisor events. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-11 03:35:35 +00:00
* Count user and OS events unless requested not to.
*/
perf_counters: allow users to count user, kernel and/or hypervisor events Impact: new perf_counter feature This extends the perf_counter_hw_event struct with bits that specify that events in user, kernel and/or hypervisor mode should not be counted (i.e. should be excluded), and adds code to program the PMU mode selection bits accordingly on x86 and powerpc. For software counters, we don't currently have the infrastructure to distinguish which mode an event occurs in, so we currently fail the counter initialization if the setting of the hw_event.exclude_* bits would require us to distinguish. Context switches and CPU migrations are currently considered to occur in kernel mode. On x86, this changes the previous policy that only root can count kernel events. Now non-root users can count kernel events or exclude them. Non-root users still can't use NMI events, though. On x86 we don't appear to have any way to control whether hypervisor events are counted or not, so hw_event.exclude_hv is ignored. On powerpc, the selection of whether to count events in user, kernel and/or hypervisor mode is PMU-wide, not per-counter, so this adds a check that the hw_event.exclude_* settings are the same as other events on the PMU. Counters being added to a group have to have the same settings as the other hardware counters in the group. Counters and groups can only be enabled in hw_perf_group_sched_in or power_perf_enable if they have the same settings as any other counters already on the PMU. If we are not running on a hypervisor, the exclude_hv setting is ignored (by forcing it to 0) since we can't ever get any hypervisor events. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-11 03:35:35 +00:00
if (!hw_event->exclude_user)
hwc->config |= ARCH_PERFMON_EVENTSEL_USR;
if (!hw_event->exclude_kernel)
hwc->config |= ARCH_PERFMON_EVENTSEL_OS;
perf_counters: allow users to count user, kernel and/or hypervisor events Impact: new perf_counter feature This extends the perf_counter_hw_event struct with bits that specify that events in user, kernel and/or hypervisor mode should not be counted (i.e. should be excluded), and adds code to program the PMU mode selection bits accordingly on x86 and powerpc. For software counters, we don't currently have the infrastructure to distinguish which mode an event occurs in, so we currently fail the counter initialization if the setting of the hw_event.exclude_* bits would require us to distinguish. Context switches and CPU migrations are currently considered to occur in kernel mode. On x86, this changes the previous policy that only root can count kernel events. Now non-root users can count kernel events or exclude them. Non-root users still can't use NMI events, though. On x86 we don't appear to have any way to control whether hypervisor events are counted or not, so hw_event.exclude_hv is ignored. On powerpc, the selection of whether to count events in user, kernel and/or hypervisor mode is PMU-wide, not per-counter, so this adds a check that the hw_event.exclude_* settings are the same as other events on the PMU. Counters being added to a group have to have the same settings as the other hardware counters in the group. Counters and groups can only be enabled in hw_perf_group_sched_in or power_perf_enable if they have the same settings as any other counters already on the PMU. If we are not running on a hypervisor, the exclude_hv setting is ignored (by forcing it to 0) since we can't ever get any hypervisor events. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-11 03:35:35 +00:00
/*
* If privileged enough, allow NMI events:
*/
hwc->nmi = 0;
if (capable(CAP_SYS_ADMIN) && hw_event->nmi)
hwc->nmi = 1;
hwc->irq_period = hw_event->irq_period;
/*
* Intel PMCs cannot be accessed sanely above 32 bit width,
* so we install an artificial 1<<31 period regardless of
* the generic counter period:
*/
if ((s64)hwc->irq_period <= 0 || hwc->irq_period > 0x7FFFFFFF)
hwc->irq_period = 0x7FFFFFFF;
atomic64_set(&hwc->period_left, hwc->irq_period);
/*
* Raw event type provide the config in the event structure
*/
if (hw_event->raw) {
hwc->config |= hw_event->type;
} else {
if (hw_event->type >= max_intel_perfmon_events)
return -EINVAL;
/*
* The generic map:
*/
hwc->config |= intel_perfmon_event_map[hw_event->type];
}
counter->wakeup_pending = 0;
return 0;
}
u64 hw_perf_save_disable(void)
{
u64 ctrl;
if (unlikely(!perf_counters_initialized))
return 0;
rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, ctrl);
wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, 0);
return ctrl;
}
EXPORT_SYMBOL_GPL(hw_perf_save_disable);
void hw_perf_restore(u64 ctrl)
{
if (unlikely(!perf_counters_initialized))
return;
wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, ctrl);
}
EXPORT_SYMBOL_GPL(hw_perf_restore);
static inline void
__pmc_fixed_disable(struct perf_counter *counter,
struct hw_perf_counter *hwc, unsigned int __idx)
{
int idx = __idx - X86_PMC_IDX_FIXED;
u64 ctrl_val, mask;
int err;
mask = 0xfULL << (idx * 4);
rdmsrl(hwc->config_base, ctrl_val);
ctrl_val &= ~mask;
err = checking_wrmsrl(hwc->config_base, ctrl_val);
}
static inline void
__pmc_generic_disable(struct perf_counter *counter,
struct hw_perf_counter *hwc, unsigned int idx)
{
if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL))
__pmc_fixed_disable(counter, hwc, idx);
else
wrmsr_safe(hwc->config_base + idx, hwc->config, 0);
}
static DEFINE_PER_CPU(u64, prev_left[X86_PMC_IDX_MAX]);
/*
* Set the next IRQ period, based on the hwc->period_left value.
* To be called with the counter disabled in hw:
*/
static void
__hw_perf_counter_set_period(struct perf_counter *counter,
struct hw_perf_counter *hwc, int idx)
{
s64 left = atomic64_read(&hwc->period_left);
s32 period = hwc->irq_period;
int err;
/*
* If we are way outside a reasoable range then just skip forward:
*/
if (unlikely(left <= -period)) {
left = period;
atomic64_set(&hwc->period_left, left);
}
if (unlikely(left <= 0)) {
left += period;
atomic64_set(&hwc->period_left, left);
}
per_cpu(prev_left[idx], smp_processor_id()) = left;
/*
* The hw counter starts counting from this counter offset,
* mark it to be able to extra future deltas:
*/
atomic64_set(&hwc->prev_count, (u64)-left);
err = checking_wrmsrl(hwc->counter_base + idx,
(u64)(-left) & counter_value_mask);
}
static inline void
__pmc_fixed_enable(struct perf_counter *counter,
struct hw_perf_counter *hwc, unsigned int __idx)
{
int idx = __idx - X86_PMC_IDX_FIXED;
u64 ctrl_val, bits, mask;
int err;
/*
perf_counters: allow users to count user, kernel and/or hypervisor events Impact: new perf_counter feature This extends the perf_counter_hw_event struct with bits that specify that events in user, kernel and/or hypervisor mode should not be counted (i.e. should be excluded), and adds code to program the PMU mode selection bits accordingly on x86 and powerpc. For software counters, we don't currently have the infrastructure to distinguish which mode an event occurs in, so we currently fail the counter initialization if the setting of the hw_event.exclude_* bits would require us to distinguish. Context switches and CPU migrations are currently considered to occur in kernel mode. On x86, this changes the previous policy that only root can count kernel events. Now non-root users can count kernel events or exclude them. Non-root users still can't use NMI events, though. On x86 we don't appear to have any way to control whether hypervisor events are counted or not, so hw_event.exclude_hv is ignored. On powerpc, the selection of whether to count events in user, kernel and/or hypervisor mode is PMU-wide, not per-counter, so this adds a check that the hw_event.exclude_* settings are the same as other events on the PMU. Counters being added to a group have to have the same settings as the other hardware counters in the group. Counters and groups can only be enabled in hw_perf_group_sched_in or power_perf_enable if they have the same settings as any other counters already on the PMU. If we are not running on a hypervisor, the exclude_hv setting is ignored (by forcing it to 0) since we can't ever get any hypervisor events. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-11 03:35:35 +00:00
* Enable IRQ generation (0x8),
* and enable ring-3 counting (0x2) and ring-0 counting (0x1)
* if requested:
*/
perf_counters: allow users to count user, kernel and/or hypervisor events Impact: new perf_counter feature This extends the perf_counter_hw_event struct with bits that specify that events in user, kernel and/or hypervisor mode should not be counted (i.e. should be excluded), and adds code to program the PMU mode selection bits accordingly on x86 and powerpc. For software counters, we don't currently have the infrastructure to distinguish which mode an event occurs in, so we currently fail the counter initialization if the setting of the hw_event.exclude_* bits would require us to distinguish. Context switches and CPU migrations are currently considered to occur in kernel mode. On x86, this changes the previous policy that only root can count kernel events. Now non-root users can count kernel events or exclude them. Non-root users still can't use NMI events, though. On x86 we don't appear to have any way to control whether hypervisor events are counted or not, so hw_event.exclude_hv is ignored. On powerpc, the selection of whether to count events in user, kernel and/or hypervisor mode is PMU-wide, not per-counter, so this adds a check that the hw_event.exclude_* settings are the same as other events on the PMU. Counters being added to a group have to have the same settings as the other hardware counters in the group. Counters and groups can only be enabled in hw_perf_group_sched_in or power_perf_enable if they have the same settings as any other counters already on the PMU. If we are not running on a hypervisor, the exclude_hv setting is ignored (by forcing it to 0) since we can't ever get any hypervisor events. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-11 03:35:35 +00:00
bits = 0x8ULL;
if (hwc->config & ARCH_PERFMON_EVENTSEL_USR)
bits |= 0x2;
if (hwc->config & ARCH_PERFMON_EVENTSEL_OS)
bits |= 0x1;
bits <<= (idx * 4);
mask = 0xfULL << (idx * 4);
rdmsrl(hwc->config_base, ctrl_val);
ctrl_val &= ~mask;
ctrl_val |= bits;
err = checking_wrmsrl(hwc->config_base, ctrl_val);
}
static void
__pmc_generic_enable(struct perf_counter *counter,
struct hw_perf_counter *hwc, int idx)
{
if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL))
__pmc_fixed_enable(counter, hwc, idx);
else
wrmsr(hwc->config_base + idx,
hwc->config | ARCH_PERFMON_EVENTSEL0_ENABLE, 0);
}
static int
fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc)
{
unsigned int event;
if (unlikely(hwc->nmi))
return -1;
event = hwc->config & ARCH_PERFMON_EVENT_MASK;
if (unlikely(event == intel_perfmon_event_map[PERF_COUNT_INSTRUCTIONS]))
return X86_PMC_IDX_FIXED_INSTRUCTIONS;
if (unlikely(event == intel_perfmon_event_map[PERF_COUNT_CPU_CYCLES]))
return X86_PMC_IDX_FIXED_CPU_CYCLES;
if (unlikely(event == intel_perfmon_event_map[PERF_COUNT_BUS_CYCLES]))
return X86_PMC_IDX_FIXED_BUS_CYCLES;
return -1;
}
/*
* Find a PMC slot for the freshly enabled / scheduled in counter:
*/
static int pmc_generic_enable(struct perf_counter *counter)
{
struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
struct hw_perf_counter *hwc = &counter->hw;
int idx;
idx = fixed_mode_idx(counter, hwc);
if (idx >= 0) {
/*
* Try to get the fixed counter, if that is already taken
* then try to get a generic counter:
*/
if (test_and_set_bit(idx, cpuc->used))
goto try_generic;
hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL;
/*
* We set it so that counter_base + idx in wrmsr/rdmsr maps to
* MSR_ARCH_PERFMON_FIXED_CTR0 ... CTR2:
*/
hwc->counter_base =
MSR_ARCH_PERFMON_FIXED_CTR0 - X86_PMC_IDX_FIXED;
hwc->idx = idx;
} else {
idx = hwc->idx;
/* Try to get the previous generic counter again */
if (test_and_set_bit(idx, cpuc->used)) {
try_generic:
idx = find_first_zero_bit(cpuc->used, nr_counters_generic);
if (idx == nr_counters_generic)
return -EAGAIN;
set_bit(idx, cpuc->used);
hwc->idx = idx;
}
hwc->config_base = MSR_ARCH_PERFMON_EVENTSEL0;
hwc->counter_base = MSR_ARCH_PERFMON_PERFCTR0;
}
perf_counters_lapic_init(hwc->nmi);
__pmc_generic_disable(counter, hwc, idx);
cpuc->counters[idx] = counter;
/*
* Make it visible before enabling the hw:
*/
smp_wmb();
__hw_perf_counter_set_period(counter, hwc, idx);
__pmc_generic_enable(counter, hwc, idx);
return 0;
}
void perf_counter_print_debug(void)
{
u64 ctrl, status, overflow, pmc_ctrl, pmc_count, prev_left, fixed;
struct cpu_hw_counters *cpuc;
int cpu, idx;
if (!nr_counters_generic)
return;
local_irq_disable();
cpu = smp_processor_id();
cpuc = &per_cpu(cpu_hw_counters, cpu);
rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, ctrl);
rdmsrl(MSR_CORE_PERF_GLOBAL_STATUS, status);
rdmsrl(MSR_CORE_PERF_GLOBAL_OVF_CTRL, overflow);
rdmsrl(MSR_ARCH_PERFMON_FIXED_CTR_CTRL, fixed);
printk(KERN_INFO "\n");
printk(KERN_INFO "CPU#%d: ctrl: %016llx\n", cpu, ctrl);
printk(KERN_INFO "CPU#%d: status: %016llx\n", cpu, status);
printk(KERN_INFO "CPU#%d: overflow: %016llx\n", cpu, overflow);
printk(KERN_INFO "CPU#%d: fixed: %016llx\n", cpu, fixed);
printk(KERN_INFO "CPU#%d: used: %016llx\n", cpu, *(u64 *)cpuc->used);
for (idx = 0; idx < nr_counters_generic; idx++) {
rdmsrl(MSR_ARCH_PERFMON_EVENTSEL0 + idx, pmc_ctrl);
rdmsrl(MSR_ARCH_PERFMON_PERFCTR0 + idx, pmc_count);
prev_left = per_cpu(prev_left[idx], cpu);
printk(KERN_INFO "CPU#%d: gen-PMC%d ctrl: %016llx\n",
cpu, idx, pmc_ctrl);
printk(KERN_INFO "CPU#%d: gen-PMC%d count: %016llx\n",
cpu, idx, pmc_count);
printk(KERN_INFO "CPU#%d: gen-PMC%d left: %016llx\n",
cpu, idx, prev_left);
}
for (idx = 0; idx < nr_counters_fixed; idx++) {
rdmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, pmc_count);
printk(KERN_INFO "CPU#%d: fixed-PMC%d count: %016llx\n",
cpu, idx, pmc_count);
}
local_irq_enable();
}
static void pmc_generic_disable(struct perf_counter *counter)
{
struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
struct hw_perf_counter *hwc = &counter->hw;
unsigned int idx = hwc->idx;
__pmc_generic_disable(counter, hwc, idx);
clear_bit(idx, cpuc->used);
cpuc->counters[idx] = NULL;
/*
* Make sure the cleared pointer becomes visible before we
* (potentially) free the counter:
*/
smp_wmb();
/*
* Drain the remaining delta count out of a counter
* that we are disabling:
*/
x86_perf_counter_update(counter, hwc, idx);
}
static void perf_store_irq_data(struct perf_counter *counter, u64 data)
{
struct perf_data *irqdata = counter->irqdata;
if (irqdata->len > PERF_DATA_BUFLEN - sizeof(u64)) {
irqdata->overrun++;
} else {
u64 *p = (u64 *) &irqdata->data[irqdata->len];
*p = data;
irqdata->len += sizeof(u64);
}
}
/*
* Save and restart an expired counter. Called by NMI contexts,
* so it has to be careful about preempting normal counter ops:
*/
static void perf_save_and_restart(struct perf_counter *counter)
{
struct hw_perf_counter *hwc = &counter->hw;
int idx = hwc->idx;
x86_perf_counter_update(counter, hwc, idx);
__hw_perf_counter_set_period(counter, hwc, idx);
if (counter->state == PERF_COUNTER_STATE_ACTIVE)
__pmc_generic_enable(counter, hwc, idx);
}
static void
perf_handle_group(struct perf_counter *sibling, u64 *status, u64 *overflown)
{
struct perf_counter *counter, *group_leader = sibling->group_leader;
/*
* Store sibling timestamps (if any):
*/
list_for_each_entry(counter, &group_leader->sibling_list, list_entry) {
x86_perf_counter_update(counter, &counter->hw, counter->hw.idx);
perf_store_irq_data(sibling, counter->hw_event.type);
perf_store_irq_data(sibling, atomic64_read(&counter->count));
}
}
/*
* Maximum interrupt frequency of 100KHz per CPU
*/
#define PERFMON_MAX_INTERRUPTS 100000/HZ
/*
* This handler is triggered by the local APIC, so the APIC IRQ handling
* rules apply:
*/
static void __smp_perf_counter_interrupt(struct pt_regs *regs, int nmi)
{
int bit, cpu = smp_processor_id();
u64 ack, status;
struct cpu_hw_counters *cpuc = &per_cpu(cpu_hw_counters, cpu);
rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, cpuc->global_enable);
/* Disable counters globally */
wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, 0);
ack_APIC_irq();
rdmsrl(MSR_CORE_PERF_GLOBAL_STATUS, status);
if (!status)
goto out;
again:
inc_irq_stat(apic_perf_irqs);
ack = status;
for_each_bit(bit, (unsigned long *)&status, X86_PMC_IDX_MAX) {
struct perf_counter *counter = cpuc->counters[bit];
clear_bit(bit, (unsigned long *) &status);
if (!counter)
continue;
perf_save_and_restart(counter);
switch (counter->hw_event.record_type) {
case PERF_RECORD_SIMPLE:
continue;
case PERF_RECORD_IRQ:
perf_store_irq_data(counter, instruction_pointer(regs));
break;
case PERF_RECORD_GROUP:
perf_handle_group(counter, &status, &ack);
break;
}
/*
* From NMI context we cannot call into the scheduler to
* do a task wakeup - but we mark these generic as
* wakeup_pending and initate a wakeup callback:
*/
if (nmi) {
counter->wakeup_pending = 1;
set_tsk_thread_flag(current, TIF_PERF_COUNTERS);
} else {
wake_up(&counter->waitq);
}
}
wrmsrl(MSR_CORE_PERF_GLOBAL_OVF_CTRL, ack);
/*
* Repeat if there is more work to be done:
*/
rdmsrl(MSR_CORE_PERF_GLOBAL_STATUS, status);
if (status)
goto again;
out:
/*
* Restore - do not reenable when global enable is off or throttled:
*/
if (++cpuc->interrupts < PERFMON_MAX_INTERRUPTS)
wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, cpuc->global_enable);
}
void perf_counter_unthrottle(void)
{
struct cpu_hw_counters *cpuc;
u64 global_enable;
if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
return;
if (unlikely(!perf_counters_initialized))
return;
cpuc = &per_cpu(cpu_hw_counters, smp_processor_id());
if (cpuc->interrupts >= PERFMON_MAX_INTERRUPTS) {
if (printk_ratelimit())
printk(KERN_WARNING "PERFMON: max interrupts exceeded!\n");
wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, cpuc->global_enable);
}
rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, global_enable);
if (unlikely(cpuc->global_enable && !global_enable))
wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, cpuc->global_enable);
cpuc->interrupts = 0;
}
void smp_perf_counter_interrupt(struct pt_regs *regs)
{
irq_enter();
apic_write(APIC_LVTPC, LOCAL_PERF_VECTOR);
__smp_perf_counter_interrupt(regs, 0);
irq_exit();
}
/*
* This handler is triggered by NMI contexts:
*/
void perf_counter_notify(struct pt_regs *regs)
{
struct cpu_hw_counters *cpuc;
unsigned long flags;
int bit, cpu;
local_irq_save(flags);
cpu = smp_processor_id();
cpuc = &per_cpu(cpu_hw_counters, cpu);
for_each_bit(bit, cpuc->used, X86_PMC_IDX_MAX) {
struct perf_counter *counter = cpuc->counters[bit];
if (!counter)
continue;
if (counter->wakeup_pending) {
counter->wakeup_pending = 0;
wake_up(&counter->waitq);
}
}
local_irq_restore(flags);
}
void perf_counters_lapic_init(int nmi)
{
u32 apic_val;
if (!perf_counters_initialized)
return;
/*
* Enable the performance counter vector in the APIC LVT:
*/
apic_val = apic_read(APIC_LVTERR);
apic_write(APIC_LVTERR, apic_val | APIC_LVT_MASKED);
if (nmi)
apic_write(APIC_LVTPC, APIC_DM_NMI);
else
apic_write(APIC_LVTPC, LOCAL_PERF_VECTOR);
apic_write(APIC_LVTERR, apic_val);
}
static int __kprobes
perf_counter_nmi_handler(struct notifier_block *self,
unsigned long cmd, void *__args)
{
struct die_args *args = __args;
struct pt_regs *regs;
if (likely(cmd != DIE_NMI_IPI))
return NOTIFY_DONE;
regs = args->regs;
apic_write(APIC_LVTPC, APIC_DM_NMI);
__smp_perf_counter_interrupt(regs, 1);
return NOTIFY_STOP;
}
static __read_mostly struct notifier_block perf_counter_nmi_notifier = {
.notifier_call = perf_counter_nmi_handler,
.next = NULL,
.priority = 1
};
void __init init_hw_perf_counters(void)
{
union cpuid10_eax eax;
unsigned int ebx;
unsigned int unused;
union cpuid10_edx edx;
if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
return;
/*
* Check whether the Architectural PerfMon supports
* Branch Misses Retired Event or not.
*/
cpuid(10, &eax.full, &ebx, &unused, &edx.full);
if (eax.split.mask_length <= ARCH_PERFMON_BRANCH_MISSES_RETIRED)
return;
printk(KERN_INFO "Intel Performance Monitoring support detected.\n");
printk(KERN_INFO "... version: %d\n", eax.split.version_id);
printk(KERN_INFO "... num counters: %d\n", eax.split.num_counters);
nr_counters_generic = eax.split.num_counters;
if (nr_counters_generic > X86_PMC_MAX_GENERIC) {
nr_counters_generic = X86_PMC_MAX_GENERIC;
WARN(1, KERN_ERR "hw perf counters %d > max(%d), clipping!",
nr_counters_generic, X86_PMC_MAX_GENERIC);
}
perf_counter_mask = (1 << nr_counters_generic) - 1;
perf_max_counters = nr_counters_generic;
printk(KERN_INFO "... bit width: %d\n", eax.split.bit_width);
counter_value_mask = (1ULL << eax.split.bit_width) - 1;
printk(KERN_INFO "... value mask: %016Lx\n", counter_value_mask);
printk(KERN_INFO "... mask length: %d\n", eax.split.mask_length);
nr_counters_fixed = edx.split.num_counters_fixed;
if (nr_counters_fixed > X86_PMC_MAX_FIXED) {
nr_counters_fixed = X86_PMC_MAX_FIXED;
WARN(1, KERN_ERR "hw perf counters fixed %d > max(%d), clipping!",
nr_counters_fixed, X86_PMC_MAX_FIXED);
}
printk(KERN_INFO "... fixed counters: %d\n", nr_counters_fixed);
perf_counter_mask |= ((1LL << nr_counters_fixed)-1) << X86_PMC_IDX_FIXED;
printk(KERN_INFO "... counter mask: %016Lx\n", perf_counter_mask);
perf_counters_initialized = true;
perf_counters_lapic_init(0);
register_die_notifier(&perf_counter_nmi_notifier);
}
static void pmc_generic_read(struct perf_counter *counter)
{
x86_perf_counter_update(counter, &counter->hw, counter->hw.idx);
}
static const struct hw_perf_counter_ops x86_perf_counter_ops = {
.enable = pmc_generic_enable,
.disable = pmc_generic_disable,
.read = pmc_generic_read,
};
const struct hw_perf_counter_ops *
hw_perf_counter_init(struct perf_counter *counter)
{
int err;
err = __hw_perf_counter_init(counter);
if (err)
return NULL;
return &x86_perf_counter_ops;
}