perf_counter: Rename perf_counter_limit sysctl
Peter Zijlstra [Thu, 11 Jun 2009 09:25:05 +0000 (11:25 +0200)]
Rename perf_counter_limit to perf_counter_max_sample_rate and
prohibit creation of counters with a known higher sample
frequency.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

include/linux/perf_counter.h
kernel/perf_counter.c
kernel/sysctl.c

index 386be91..95c797c 100644 (file)
@@ -650,7 +650,7 @@ extern struct perf_callchain_entry *perf_callchain(struct pt_regs *regs);
 
 extern int sysctl_perf_counter_paranoid;
 extern int sysctl_perf_counter_mlock;
-extern int sysctl_perf_counter_limit;
+extern int sysctl_perf_counter_sample_rate;
 
 extern void perf_counter_init(void);
 
index 63f1987..3b2829d 100644 (file)
@@ -44,11 +44,12 @@ static atomic_t nr_mmap_counters __read_mostly;
 static atomic_t nr_comm_counters __read_mostly;
 
 /*
- * 0 - not paranoid
- * 1 - disallow cpu counters to unpriv
- * 2 - disallow kernel profiling to unpriv
+ * perf counter paranoia level:
+ *  0 - not paranoid
+ *  1 - disallow cpu counters to unpriv
+ *  2 - disallow kernel profiling to unpriv
  */
-int sysctl_perf_counter_paranoid __read_mostly; /* do we need to be privileged */
+int sysctl_perf_counter_paranoid __read_mostly;
 
 static inline bool perf_paranoid_cpu(void)
 {
@@ -61,7 +62,11 @@ static inline bool perf_paranoid_kernel(void)
 }
 
 int sysctl_perf_counter_mlock __read_mostly = 512; /* 'free' kb per user */
-int sysctl_perf_counter_limit __read_mostly = 100000; /* max NMIs per second */
+
+/*
+ * max perf counter sample rate
+ */
+int sysctl_perf_counter_sample_rate __read_mostly = 100000;
 
 static atomic64_t perf_counter_id;
 
@@ -1244,7 +1249,7 @@ static void perf_ctx_adjust_freq(struct perf_counter_context *ctx)
                if (interrupts == MAX_INTERRUPTS) {
                        perf_log_throttle(counter, 1);
                        counter->pmu->unthrottle(counter);
-                       interrupts = 2*sysctl_perf_counter_limit/HZ;
+                       interrupts = 2*sysctl_perf_counter_sample_rate/HZ;
                }
 
                if (!counter->attr.freq || !counter->attr.sample_freq)
@@ -1682,7 +1687,7 @@ static int perf_counter_period(struct perf_counter *counter, u64 __user *arg)
 
        spin_lock_irq(&ctx->lock);
        if (counter->attr.freq) {
-               if (value > sysctl_perf_counter_limit) {
+               if (value > sysctl_perf_counter_sample_rate) {
                        ret = -EINVAL;
                        goto unlock;
                }
@@ -2979,7 +2984,8 @@ int perf_counter_overflow(struct perf_counter *counter, int nmi,
        } else {
                if (hwc->interrupts != MAX_INTERRUPTS) {
                        hwc->interrupts++;
-                       if (HZ * hwc->interrupts > (u64)sysctl_perf_counter_limit) {
+                       if (HZ * hwc->interrupts >
+                                       (u64)sysctl_perf_counter_sample_rate) {
                                hwc->interrupts = MAX_INTERRUPTS;
                                perf_log_throttle(counter, 0);
                                ret = 1;
@@ -3639,6 +3645,11 @@ SYSCALL_DEFINE5(perf_counter_open,
                        return -EACCES;
        }
 
+       if (attr.freq) {
+               if (attr.sample_freq > sysctl_perf_counter_sample_rate)
+                       return -EINVAL;
+       }
+
        /*
         * Get the target context (task or percpu):
         */
index 344a659..9fd4e43 100644 (file)
@@ -932,9 +932,9 @@ static struct ctl_table kern_table[] = {
        },
        {
                .ctl_name       = CTL_UNNUMBERED,
-               .procname       = "perf_counter_int_limit",
-               .data           = &sysctl_perf_counter_limit,
-               .maxlen         = sizeof(sysctl_perf_counter_limit),
+               .procname       = "perf_counter_max_sample_rate",
+               .data           = &sysctl_perf_counter_sample_rate,
+               .maxlen         = sizeof(sysctl_perf_counter_sample_rate),
                .mode           = 0644,
                .proc_handler   = &proc_dointvec,
        },