Merge branch 'perf/fast' into perf/core

Merge reason: Lets ready it for v3.4
Signed-off-by: Ingo Molnar <mingo@elte.hu>

arch/arm/include/asm/perf_event.h

@@ -12,10 +12,6 @@
 #ifndef __ARM_PERF_EVENT_H__
 #define __ARM_PERF_EVENT_H__
 
-/* ARM performance counters start from 1 (in the cp15 accesses) so use the
- * same indexes here for consistency. */
-#define PERF_EVENT_INDEX_OFFSET 1
-
 /* ARM perf PMU IDs for use by internal perf clients. */
 enum arm_perf_pmu_ids {
 	ARM_PERF_PMU_ID_XSCALE1	= 0,

arch/frv/include/asm/perf_event.h

@@ -12,6 +12,4 @@
 #ifndef _ASM_PERF_EVENT_H
 #define _ASM_PERF_EVENT_H
 
-#define PERF_EVENT_INDEX_OFFSET	0
-
 #endif /* _ASM_PERF_EVENT_H */

arch/hexagon/include/asm/perf_event.h

@@ -19,6 +19,4 @@
 #ifndef _ASM_PERF_EVENT_H
 #define _ASM_PERF_EVENT_H
 
-#define PERF_EVENT_INDEX_OFFSET	0
-
 #endif /* _ASM_PERF_EVENT_H */

arch/powerpc/include/asm/perf_event_server.h

@@ -61,8 +61,6 @@ struct pt_regs;
 extern unsigned long perf_misc_flags(struct pt_regs *regs);
 extern unsigned long perf_instruction_pointer(struct pt_regs *regs);
 
-#define PERF_EVENT_INDEX_OFFSET	1
-
 /*
  * Only override the default definitions in include/linux/perf_event.h
  * if we have hardware PMU support.

arch/powerpc/kernel/perf_event.c

@@ -1187,6 +1187,11 @@ static int power_pmu_event_init(struct perf_event *event)
 	return err;
 }
 
+static int power_pmu_event_idx(struct perf_event *event)
+{
+	return event->hw.idx;
+}
+
 struct pmu power_pmu = {
 	.pmu_enable	= power_pmu_enable,
 	.pmu_disable	= power_pmu_disable,
@@ -1199,6 +1204,7 @@ struct pmu power_pmu = {
 	.start_txn	= power_pmu_start_txn,
 	.cancel_txn	= power_pmu_cancel_txn,
 	.commit_txn	= power_pmu_commit_txn,
+	.event_idx	= power_pmu_event_idx,
 };
 
 /*

arch/s390/include/asm/perf_event.h

@@ -6,4 +6,3 @@
 
 /* Empty, just to avoid compiling error */
 
-#define PERF_EVENT_INDEX_OFFSET 0

arch/x86/include/asm/perf_event.h

@@ -188,8 +188,6 @@ extern u32 get_ibs_caps(void);
 #ifdef CONFIG_PERF_EVENTS
 extern void perf_events_lapic_init(void);
 
-#define PERF_EVENT_INDEX_OFFSET			0
-
 /*
  * Abuse bit 3 of the cpu eflags register to indicate proper PEBS IP fixups.
  * This flag is otherwise unused and ABI specified to be 0, so nobody should

arch/x86/kernel/cpu/perf_event.c

@@ -24,6 +24,7 @@
 #include <linux/slab.h>
 #include <linux/cpu.h>
 #include <linux/bitops.h>
+#include <linux/device.h>
 
 #include <asm/apic.h>
 #include <asm/stacktrace.h>
@@ -31,6 +32,7 @@
 #include <asm/compat.h>
 #include <asm/smp.h>
 #include <asm/alternative.h>
+#include <asm/timer.h>
 
 #include "perf_event.h"
 
@@ -1210,6 +1212,8 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
 		break;
 
 	case CPU_STARTING:
+		if (x86_pmu.attr_rdpmc)
+			set_in_cr4(X86_CR4_PCE);
 		if (x86_pmu.cpu_starting)
 			x86_pmu.cpu_starting(cpu);
 		break;
@@ -1319,6 +1323,8 @@ static int __init init_hw_perf_events(void)
 		}
 	}
 
+	x86_pmu.attr_rdpmc = 1; /* enable userspace RDPMC usage by default */
+
 	pr_info("... version:                %d\n",     x86_pmu.version);
 	pr_info("... bit width:              %d\n",     x86_pmu.cntval_bits);
 	pr_info("... generic registers:      %d\n",     x86_pmu.num_counters);
@@ -1542,10 +1548,71 @@ static int x86_pmu_event_init(struct perf_event *event)
 	return err;
 }
 
+static int x86_pmu_event_idx(struct perf_event *event)
+{
+	int idx = event->hw.idx;
+
+	if (x86_pmu.num_counters_fixed && idx >= X86_PMC_IDX_FIXED) {
+		idx -= X86_PMC_IDX_FIXED;
+		idx |= 1 << 30;
+	}
+
+	return idx + 1;
+}
+
+static ssize_t get_attr_rdpmc(struct device *cdev,
+			      struct device_attribute *attr,
+			      char *buf)
+{
+	return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc);
+}
+
+static void change_rdpmc(void *info)
+{
+	bool enable = !!(unsigned long)info;
+
+	if (enable)
+		set_in_cr4(X86_CR4_PCE);
+	else
+		clear_in_cr4(X86_CR4_PCE);
+}
+
+static ssize_t set_attr_rdpmc(struct device *cdev,
+			      struct device_attribute *attr,
+			      const char *buf, size_t count)
+{
+	unsigned long val = simple_strtoul(buf, NULL, 0);
+
+	if (!!val != !!x86_pmu.attr_rdpmc) {
+		x86_pmu.attr_rdpmc = !!val;
+		smp_call_function(change_rdpmc, (void *)val, 1);
+	}
+
+	return count;
+}
+
+static DEVICE_ATTR(rdpmc, S_IRUSR | S_IWUSR, get_attr_rdpmc, set_attr_rdpmc);
+
+static struct attribute *x86_pmu_attrs[] = {
+	&dev_attr_rdpmc.attr,
+	NULL,
+};
+
+static struct attribute_group x86_pmu_attr_group = {
+	.attrs = x86_pmu_attrs,
+};
+
+static const struct attribute_group *x86_pmu_attr_groups[] = {
+	&x86_pmu_attr_group,
+	NULL,
+};
+
 static struct pmu pmu = {
 	.pmu_enable	= x86_pmu_enable,
 	.pmu_disable	= x86_pmu_disable,
 
+	.attr_groups	= x86_pmu_attr_groups,
+
 	.event_init	= x86_pmu_event_init,
 
 	.add		= x86_pmu_add,
@@ -1557,8 +1624,23 @@ static struct pmu pmu = {
 	.start_txn	= x86_pmu_start_txn,
 	.cancel_txn	= x86_pmu_cancel_txn,
 	.commit_txn	= x86_pmu_commit_txn,
+
+	.event_idx	= x86_pmu_event_idx,
 };
 
+void perf_update_user_clock(struct perf_event_mmap_page *userpg, u64 now)
+{
+	if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
+		return;
+
+	if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
+		return;
+
+	userpg->time_mult = this_cpu_read(cyc2ns);
+	userpg->time_shift = CYC2NS_SCALE_FACTOR;
+	userpg->time_offset = this_cpu_read(cyc2ns_offset) - now;
+}
+
 /*
  * callchain support
  */

arch/x86/kernel/cpu/perf_event.h

@@ -307,6 +307,14 @@ struct x86_pmu {
 	struct x86_pmu_quirk *quirks;
 	int		perfctr_second_write;
 
+	/*
+	 * sysfs attrs
+	 */
+	int		attr_rdpmc;
+
+	/*
+	 * CPU Hotplug hooks
+	 */
 	int		(*cpu_prepare)(int cpu);
 	void		(*cpu_starting)(int cpu);
 	void		(*cpu_dying)(int cpu);

include/linux/perf_event.h

@@ -291,12 +291,14 @@ struct perf_event_mmap_page {
 	__s64	offset;			/* add to hardware event value */
 	__u64	time_enabled;		/* time event active */
 	__u64	time_running;		/* time event on cpu */
+	__u32	time_mult, time_shift;
+	__u64	time_offset;
 
 		/*
 		 * Hole for extension of the self monitor capabilities
 		 */
 
-	__u64	__reserved[123];	/* align to 1k */
+	__u64	__reserved[121];	/* align to 1k */
 
 	/*
 	 * Control data for the mmap() data buffer.
@@ -615,6 +617,7 @@ struct pmu {
 	struct list_head		entry;
 
 	struct device			*dev;
+	const struct attribute_group	**attr_groups;
 	char				*name;
 	int				type;
 
@@ -680,6 +683,12 @@ struct pmu {
 	 * for each successful ->add() during the transaction.
 	 */
 	void (*cancel_txn)		(struct pmu *pmu); /* optional */
+
+	/*
+	 * Will return the value for perf_event_mmap_page::index for this event,
+	 * if no implementation is provided it will default to: event->hw.idx + 1.
+	 */
+	int (*event_idx)		(struct perf_event *event); /*optional */
 };
 
 /**

kernel/events/core.c

@@ -3208,10 +3208,6 @@ int perf_event_task_disable(void)
 	return 0;
 }
 
-#ifndef PERF_EVENT_INDEX_OFFSET
-# define PERF_EVENT_INDEX_OFFSET 0
-#endif
-
 static int perf_event_index(struct perf_event *event)
 {
 	if (event->hw.state & PERF_HES_STOPPED)
@@ -3220,21 +3216,26 @@ static int perf_event_index(struct perf_event *event)
 	if (event->state != PERF_EVENT_STATE_ACTIVE)
 		return 0;
 
-	return event->hw.idx + 1 - PERF_EVENT_INDEX_OFFSET;
+	return event->pmu->event_idx(event);
 }
 
 static void calc_timer_values(struct perf_event *event,
+				u64 *now,
 				u64 *enabled,
 				u64 *running)
 {
-	u64 now, ctx_time;
+	u64 ctx_time;
 
-	now = perf_clock();
-	ctx_time = event->shadow_ctx_time + now;
+	*now = perf_clock();
+	ctx_time = event->shadow_ctx_time + *now;
 	*enabled = ctx_time - event->tstamp_enabled;
 	*running = ctx_time - event->tstamp_running;
 }
 
+void __weak perf_update_user_clock(struct perf_event_mmap_page *userpg, u64 now)
+{
+}
+
 /*
  * Callers need to ensure there can be no nesting of this function, otherwise
  * the seqlock logic goes bad. We can not serialize this because the arch
@@ -3244,7 +3245,7 @@ void perf_event_update_userpage(struct perf_event *event)
 {
 	struct perf_event_mmap_page *userpg;
 	struct ring_buffer *rb;
-	u64 enabled, running;
+	u64 enabled, running, now;
 
 	rcu_read_lock();
 	/*
@@ -3256,7 +3257,7 @@ void perf_event_update_userpage(struct perf_event *event)
 	 * because of locking issue as we can be called in
 	 * NMI context
 	 */
-	calc_timer_values(event, &enabled, &running);
+	calc_timer_values(event, &now, &enabled, &running);
 	rb = rcu_dereference(event->rb);
 	if (!rb)
 		goto unlock;
@@ -3272,7 +3273,7 @@ void perf_event_update_userpage(struct perf_event *event)
 	barrier();
 	userpg->index = perf_event_index(event);
 	userpg->offset = perf_event_count(event);
-	if (event->state == PERF_EVENT_STATE_ACTIVE)
+	if (userpg->index)
 		userpg->offset -= local64_read(&event->hw.prev_count);
 
 	userpg->time_enabled = enabled +
@@ -3281,6 +3282,8 @@ void perf_event_update_userpage(struct perf_event *event)
 	userpg->time_running = running +
 			atomic64_read(&event->child_total_time_running);
 
+	perf_update_user_clock(userpg, now);
+
 	barrier();
 	++userpg->lock;
 	preempt_enable();
@@ -3538,6 +3541,8 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma)
 	event->mmap_user = get_current_user();
 	vma->vm_mm->pinned_vm += event->mmap_locked;
 
+	perf_event_update_userpage(event);
+
 unlock:
 	if (!ret)
 		atomic_inc(&event->mmap_count);
@@ -3769,7 +3774,7 @@ static void perf_output_read_group(struct perf_output_handle *handle,
 static void perf_output_read(struct perf_output_handle *handle,
 			     struct perf_event *event)
 {
-	u64 enabled = 0, running = 0;
+	u64 enabled = 0, running = 0, now;
 	u64 read_format = event->attr.read_format;
 
 	/*
@@ -3782,7 +3787,7 @@ static void perf_output_read(struct perf_output_handle *handle,
 	 * NMI context
 	 */
 	if (read_format & PERF_FORMAT_TOTAL_TIMES)
-		calc_timer_values(event, &enabled, &running);
+		calc_timer_values(event, &now, &enabled, &running);
 
 	if (event->attr.read_format & PERF_FORMAT_GROUP)
 		perf_output_read_group(handle, event, enabled, running);
@@ -4994,6 +4999,11 @@ static int perf_swevent_init(struct perf_event *event)
 	return 0;
 }
 
+static int perf_swevent_event_idx(struct perf_event *event)
+{
+	return 0;
+}
+
 static struct pmu perf_swevent = {
 	.task_ctx_nr	= perf_sw_context,
 
@@ -5003,6 +5013,8 @@ static struct pmu perf_swevent = {
 	.start		= perf_swevent_start,
 	.stop		= perf_swevent_stop,
 	.read		= perf_swevent_read,
+
+	.event_idx	= perf_swevent_event_idx,
 };
 
 #ifdef CONFIG_EVENT_TRACING
@@ -5089,6 +5101,8 @@ static struct pmu perf_tracepoint = {
 	.start		= perf_swevent_start,
 	.stop		= perf_swevent_stop,
 	.read		= perf_swevent_read,
+
+	.event_idx	= perf_swevent_event_idx,
 };
 
 static inline void perf_tp_register(void)
@@ -5308,6 +5322,8 @@ static struct pmu perf_cpu_clock = {
 	.start		= cpu_clock_event_start,
 	.stop		= cpu_clock_event_stop,
 	.read		= cpu_clock_event_read,
+
+	.event_idx	= perf_swevent_event_idx,
 };
 
 /*
@@ -5380,6 +5396,8 @@ static struct pmu perf_task_clock = {
 	.start		= task_clock_event_start,
 	.stop		= task_clock_event_stop,
 	.read		= task_clock_event_read,
+
+	.event_idx	= perf_swevent_event_idx,
 };
 
 static void perf_pmu_nop_void(struct pmu *pmu)
@@ -5407,6 +5425,11 @@ static void perf_pmu_cancel_txn(struct pmu *pmu)
 	perf_pmu_enable(pmu);
 }
 
+static int perf_event_idx_default(struct perf_event *event)
+{
+	return event->hw.idx + 1;
+}
+
 /*
  * Ensures all contexts with the same task_ctx_nr have the same
  * pmu_cpu_context too.
@@ -5493,6 +5516,7 @@ static int pmu_dev_alloc(struct pmu *pmu)
 	if (!pmu->dev)
 		goto out;
 
+	pmu->dev->groups = pmu->attr_groups;
 	device_initialize(pmu->dev);
 	ret = dev_set_name(pmu->dev, "%s", pmu->name);
 	if (ret)
@@ -5596,6 +5620,9 @@ int perf_pmu_register(struct pmu *pmu, char *name, int type)
 		pmu->pmu_disable = perf_pmu_nop_void;
 	}
 
+	if (!pmu->event_idx)
+		pmu->event_idx = perf_event_idx_default;
+
 	list_add_rcu(&pmu->entry, &pmus);
 	ret = 0;
 unlock:

kernel/events/hw_breakpoint.c

@@ -613,6 +613,11 @@ static void hw_breakpoint_stop(struct perf_event *bp, int flags)
 	bp->hw.state = PERF_HES_STOPPED;
 }
 
+static int hw_breakpoint_event_idx(struct perf_event *bp)
+{
+	return 0;
+}
+
 static struct pmu perf_breakpoint = {
 	.task_ctx_nr	= perf_sw_context, /* could eventually get its own */
 
@@ -622,6 +627,8 @@ static struct pmu perf_breakpoint = {
 	.start		= hw_breakpoint_start,
 	.stop		= hw_breakpoint_stop,
 	.read		= hw_breakpoint_pmu_read,
+
+	.event_idx	= hw_breakpoint_event_idx,
 };
 
 int __init init_hw_breakpoint(void)

tools/perf/builtin-test.c

@@ -15,6 +15,8 @@
 #include "util/thread_map.h"
 #include "../../include/linux/hw_breakpoint.h"
 
+#include <sys/mman.h>
+
 static int vmlinux_matches_kallsyms_filter(struct map *map __used, struct symbol *sym)
 {
 	bool *visited = symbol__priv(sym);
@@ -1296,6 +1298,173 @@ static int test__PERF_RECORD(void)
 	return (err < 0 || errs > 0) ? -1 : 0;
 }
 
+
+#if defined(__x86_64__) || defined(__i386__)
+
+#define barrier() asm volatile("" ::: "memory")
+
+static u64 rdpmc(unsigned int counter)
+{
+	unsigned int low, high;
+
+	asm volatile("rdpmc" : "=a" (low), "=d" (high) : "c" (counter));
+
+	return low | ((u64)high) << 32;
+}
+
+static u64 rdtsc(void)
+{
+	unsigned int low, high;
+
+	asm volatile("rdtsc" : "=a" (low), "=d" (high));
+
+	return low | ((u64)high) << 32;
+}
+
+static u64 mmap_read_self(void *addr)
+{
+	struct perf_event_mmap_page *pc = addr;
+	u32 seq, idx, time_mult = 0, time_shift = 0;
+	u64 count, cyc = 0, time_offset = 0, enabled, running, delta;
+
+	do {
+		seq = pc->lock;
+		barrier();
+
+		enabled = pc->time_enabled;
+		running = pc->time_running;
+
+		if (enabled != running) {
+			cyc = rdtsc();
+			time_mult = pc->time_mult;
+			time_shift = pc->time_shift;
+			time_offset = pc->time_offset;
+		}
+
+		idx = pc->index;
+		count = pc->offset;
+		if (idx)
+			count += rdpmc(idx - 1);
+
+		barrier();
+	} while (pc->lock != seq);
+
+	if (enabled != running) {
+		u64 quot, rem;
+
+		quot = (cyc >> time_shift);
+		rem = cyc & ((1 << time_shift) - 1);
+		delta = time_offset + quot * time_mult +
+			((rem * time_mult) >> time_shift);
+
+		enabled += delta;
+		if (idx)
+			running += delta;
+
+		quot = count / running;
+		rem = count % running;
+		count = quot * enabled + (rem * enabled) / running;
+	}
+
+	return count;
+}
+
+/*
+ * If the RDPMC instruction faults then signal this back to the test parent task:
+ */
+static void segfault_handler(int sig __used, siginfo_t *info __used, void *uc __used)
+{
+	exit(-1);
+}
+
+static int __test__rdpmc(void)
+{
+	long page_size = sysconf(_SC_PAGE_SIZE);
+	volatile int tmp = 0;
+	u64 i, loops = 1000;
+	int n;
+	int fd;
+	void *addr;
+	struct perf_event_attr attr = {
+		.type = PERF_TYPE_HARDWARE,
+		.config = PERF_COUNT_HW_INSTRUCTIONS,
+		.exclude_kernel = 1,
+	};
+	u64 delta_sum = 0;
+        struct sigaction sa;
+
+	sigfillset(&sa.sa_mask);
+	sa.sa_sigaction = segfault_handler;
+	sigaction(SIGSEGV, &sa, NULL);
+
+	fprintf(stderr, "\n\n");
+
+	fd = sys_perf_event_open(&attr, 0, -1, -1, 0);
+	if (fd < 0) {
+		die("Error: sys_perf_event_open() syscall returned "
+		    "with %d (%s)\n", fd, strerror(errno));
+	}
+
+	addr = mmap(NULL, page_size, PROT_READ, MAP_SHARED, fd, 0);
+	if (addr == (void *)(-1)) {
+		die("Error: mmap() syscall returned "
+		    "with (%s)\n", strerror(errno));
+	}
+
+	for (n = 0; n < 6; n++) {
+		u64 stamp, now, delta;
+
+		stamp = mmap_read_self(addr);
+
+		for (i = 0; i < loops; i++)
+			tmp++;
+
+		now = mmap_read_self(addr);
+		loops *= 10;
+
+		delta = now - stamp;
+		fprintf(stderr, "%14d: %14Lu\n", n, (long long)delta);
+
+		delta_sum += delta;
+	}
+
+	munmap(addr, page_size);
+	close(fd);
+
+	fprintf(stderr, "   ");
+
+	if (!delta_sum)
+		return -1;
+
+	return 0;
+}
+
+static int test__rdpmc(void)
+{
+	int status = 0;
+	int wret = 0;
+	int ret;
+	int pid;
+
+	pid = fork();
+	if (pid < 0)
+		return -1;
+
+	if (!pid) {
+		ret = __test__rdpmc();
+
+		exit(ret);
+	}
+
+	wret = waitpid(pid, &status, 0);
+	if (wret < 0 || status)
+		return -1;
+
+	return 0;
+}
+
+#endif
+
 static struct test {
 	const char *desc;
 	int (*func)(void);
@@ -1320,6 +1489,12 @@ static struct test {
 		.desc = "parse events tests",
 		.func = test__parse_events,
 	},
+#if defined(__x86_64__) || defined(__i386__)
+	{
+		.desc = "x86 rdpmc test",
+		.func = test__rdpmc,
+	},
+#endif
 	{
 		.desc = "Validate PERF_RECORD_* events & perf_sample fields",
 		.func = test__PERF_RECORD,
@@ -1412,7 +1587,5 @@ int cmd_test(int argc, const char **argv, const char *prefix __used)
 	if (symbol__init() < 0)
 		return -1;
 
-	setup_pager();
-
 	return __cmd_test(argc, argv);
 }