perf_counter, x86: implement the interrupt handler for AMD cpus

This patch implements the interrupt handler for AMD performance
counters. In difference to the Intel pmu, there is no single status
register and also there are no fixed counters. This makes the handler
very different and it is useful to make the handler vendor
specific. To check if a counter is overflowed the upper bit of the
counter is checked. Only counters where the active bit is set are
checked.

With this patch throttling is enabled for AMD performance counters.

This patch also reenables Linux performance counters on AMD cpus.

[ Impact: re-enable perfcounters on AMD CPUs ]

Signed-off-by: Robert Richter <robert.richter@amd.com>
Cc: Paul Mackerras <paulus@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1241002046-8832-25-git-send-email-robert.richter@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c
index 2d3681bbb5225429608b70811c18d97aca780344..f4d59d4cf3f1878af849c9d82bcd157f2924eef5 100644 (file)
--- a/arch/x86/kernel/cpu/perf_counter.c
+++ b/arch/x86/kernel/cpu/perf_counter.c
@@ -240,10 +240,6 @@ static int __hw_perf_counter_init(struct perf_counter *counter)
        struct hw_perf_counter *hwc = &counter->hw;
        int err;
 
-       /* disable temporarily */
-       if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
-               return -ENOSYS;
-
        if (!x86_pmu_initialized())
                return -ENODEV;
 
@@ -773,7 +769,43 @@ out:
        return ret;
 }
 
-static int amd_pmu_handle_irq(struct pt_regs *regs, int nmi) { return 0; }
+static int amd_pmu_handle_irq(struct pt_regs *regs, int nmi)
+{
+       int cpu = smp_processor_id();
+       struct cpu_hw_counters *cpuc = &per_cpu(cpu_hw_counters, cpu);
+       u64 val;
+       int handled = 0;
+       struct perf_counter *counter;
+       struct hw_perf_counter *hwc;
+       int idx;
+
+       ++cpuc->interrupts;
+       for (idx = 0; idx < x86_pmu.num_counters; idx++) {
+               if (!test_bit(idx, cpuc->active))
+                       continue;
+               counter = cpuc->counters[idx];
+               hwc = &counter->hw;
+               x86_perf_counter_update(counter, hwc, idx);
+               val = atomic64_read(&hwc->prev_count);
+               if (val & (1ULL << (x86_pmu.counter_bits - 1)))
+                       continue;
+               /* counter overflow */
+               x86_perf_counter_set_period(counter, hwc, idx);
+               handled = 1;
+               inc_irq_stat(apic_perf_irqs);
+               if (perf_counter_overflow(counter, nmi, regs, 0))
+                       amd_pmu_disable_counter(hwc, idx);
+               else if (cpuc->interrupts >= PERFMON_MAX_INTERRUPTS)
+                       /*
+                        * do not reenable when throttled, but reload
+                        * the register
+                        */
+                       amd_pmu_disable_counter(hwc, idx);
+               else if (counter->state == PERF_COUNTER_STATE_ACTIVE)
+                       amd_pmu_enable_counter(hwc, idx);
+       }
+       return handled;
+}
 
 void perf_counter_unthrottle(void)
 {
@@ -782,9 +814,6 @@ void perf_counter_unthrottle(void)
        if (!x86_pmu_initialized())
                return;
 
-       if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
-               return;
-
        cpuc = &__get_cpu_var(cpu_hw_counters);
        if (cpuc->interrupts >= PERFMON_MAX_INTERRUPTS) {
                if (printk_ratelimit())