perf counters: restructure the API

Impact: clean up new API

Thorough cleanup of the new perf counters API, we now get clean separation
of the various concepts:

 - introduce perf_counter_hw_event to separate out the event source details

 - move special type flags into separate attributes: PERF_COUNT_NMI,
   PERF_COUNT_RAW

 - extend the type to u64 and reserve it fully to the architecture in the
   raw type case.

And make use of all these changes in the core and x86 perfcounters code.

Also change the syscall signature to:

  asmlinkage int sys_perf_counter_open(

	struct perf_counter_hw_event	*hw_event_uptr		__user,
	pid_t				pid,
	int				cpu,
	int				group_fd);

( Note that group_fd is unused for now - it's reserved for the counter
  groups abstraction. )
Signed-off-by: Ingo Molnar <mingo@elte.hu>

arch/x86/kernel/cpu/perf_counter.c

@@ -58,8 +58,8 @@ const int max_intel_perfmon_events = ARRAY_SIZE(intel_perfmon_event_map);
  */
 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;
-	u32 hw_event_type = counter->event.hw_event_type;
 
 	if (unlikely(!perf_counters_initialized))
 		return -EINVAL;
@@ -77,14 +77,14 @@ int hw_perf_counter_init(struct perf_counter *counter)
 	hwc->nmi = 0;
 	if (capable(CAP_SYS_ADMIN)) {
 		hwc->config |= ARCH_PERFMON_EVENTSEL_OS;
-		if (hw_event_type & PERF_COUNT_NMI)
+		if (hw_event->nmi)
 			hwc->nmi = 1;
 	}
 
-	hwc->config_base = MSR_ARCH_PERFMON_EVENTSEL0;
-	hwc->counter_base = MSR_ARCH_PERFMON_PERFCTR0;
+	hwc->config_base	= MSR_ARCH_PERFMON_EVENTSEL0;
+	hwc->counter_base	= MSR_ARCH_PERFMON_PERFCTR0;
 
-	hwc->irq_period = counter->event.hw_event_period;
+	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
@@ -93,21 +93,20 @@ int hw_perf_counter_init(struct perf_counter *counter)
 	if (!hwc->irq_period)
 		hwc->irq_period = 0x7FFFFFFF;
 
-	hwc->next_count = -((s32) hwc->irq_period);
+	hwc->next_count	= -(s32)hwc->irq_period;
 
 	/*
 	 * Raw event type provide the config in the event structure
 	 */
-	hw_event_type &= ~PERF_COUNT_NMI;
-	if (hw_event_type == PERF_COUNT_RAW) {
-		hwc->config |= counter->event.hw_raw_ctrl;
+	if (hw_event->raw) {
+		hwc->config |= hw_event->type;
 	} else {
-		if (hw_event_type >= max_intel_perfmon_events)
+		if (hw_event->type >= max_intel_perfmon_events)
 			return -EINVAL;
 		/*
 		 * The generic map:
 		 */
-		hwc->config |= intel_perfmon_event_map[hw_event_type];
+		hwc->config |= intel_perfmon_event_map[hw_event->type];
 	}
 	counter->wakeup_pending = 0;
 
@@ -354,7 +353,7 @@ perf_handle_group(struct perf_counter *leader, u64 *status, u64 *overflown)
 	int bit;
 
 	list_for_each_entry(counter, &ctx->counters, list) {
-		if (counter->record_type != PERF_RECORD_SIMPLE ||
+		if (counter->hw_event.record_type != PERF_RECORD_SIMPLE ||
 		    counter == leader)
 			continue;
 
@@ -372,7 +371,7 @@ perf_handle_group(struct perf_counter *leader, u64 *status, u64 *overflown)
 				perf_save_and_restart(counter);
 			}
 		}
-		perf_store_irq_data(leader, counter->event.hw_event_type);
+		perf_store_irq_data(leader, counter->hw_event.type);
 		perf_store_irq_data(leader, atomic64_counter_read(counter));
 	}
 }
@@ -410,7 +409,7 @@ again:
 
 		perf_save_and_restart(counter);
 
-		switch (counter->record_type) {
+		switch (counter->hw_event.record_type) {
 		case PERF_RECORD_SIMPLE:
 			continue;
 		case PERF_RECORD_IRQ:
@@ -418,7 +417,7 @@ again:
 			break;
 		case PERF_RECORD_GROUP:
 			perf_store_irq_data(counter,
-					    counter->event.hw_event_type);
+					    counter->hw_event.type);
 			perf_store_irq_data(counter,
 					    atomic64_counter_read(counter));
 			perf_handle_group(counter, &status, &ack);

include/linux/perf_counter.h

@@ -24,65 +24,93 @@
 struct task_struct;
 
 /*
- * Generalized hardware event types, used by the hw_event_type parameter
- * of the sys_perf_counter_open() syscall:
+ * User-space ABI bits:
+ */
+
+/*
+ * Generalized performance counter event types, used by the hw_event.type
+ * parameter of the sys_perf_counter_open() syscall:
  */
 enum hw_event_types {
-	PERF_COUNT_CYCLES,
-	PERF_COUNT_INSTRUCTIONS,
-	PERF_COUNT_CACHE_REFERENCES,
-	PERF_COUNT_CACHE_MISSES,
-	PERF_COUNT_BRANCH_INSTRUCTIONS,
-	PERF_COUNT_BRANCH_MISSES,
 	/*
-	 * If this bit is set in the type, then trigger NMI sampling:
+	 * Common hardware events, generalized by the kernel:
 	 */
-	PERF_COUNT_NMI			= (1 << 30),
-	PERF_COUNT_RAW			= (1 << 31),
+	PERF_COUNT_CYCLES		=  0,
+	PERF_COUNT_INSTRUCTIONS		=  1,
+	PERF_COUNT_CACHE_REFERENCES	=  2,
+	PERF_COUNT_CACHE_MISSES		=  3,
+	PERF_COUNT_BRANCH_INSTRUCTIONS	=  4,
+	PERF_COUNT_BRANCH_MISSES	=  5,
+
+	/*
+	 * Special "software" counters provided by the kernel, even if
+	 * the hardware does not support performance counters. These
+	 * counters measure various physical and sw events of the
+	 * kernel (and allow the profiling of them as well):
+	 */
+	PERF_COUNT_CPU_CLOCK		= -1,
+	PERF_COUNT_TASK_CLOCK		= -2,
+	PERF_COUNT_PAGE_FAULTS		= -3,
+	PERF_COUNT_CONTEXT_SWITCHES	= -4,
 };
 
 /*
  * IRQ-notification data record type:
  */
-enum perf_record_type {
-	PERF_RECORD_SIMPLE,
-	PERF_RECORD_IRQ,
-	PERF_RECORD_GROUP,
+enum perf_counter_record_type {
+	PERF_RECORD_SIMPLE		=  0,
+	PERF_RECORD_IRQ			=  1,
+	PERF_RECORD_GROUP		=  2,
 };
 
-struct perf_counter_event {
-	u32			hw_event_type;
-	u32			hw_event_period;
-	u64			hw_raw_ctrl;
+/*
+ * Hardware event to monitor via a performance monitoring counter:
+ */
+struct perf_counter_hw_event {
+	u64			type;
+
+	u64			irq_period;
+	u32			record_type;
+
+	u32			disabled     :  1, /* off by default */
+				nmi	     :  1, /* NMI sampling   */
+				raw	     :  1, /* raw event type */
+				__reserved_1 : 29;
+
+	u64			__reserved_2;
 };
 
+/*
+ * Kernel-internal data types:
+ */
+
 /**
- * struct hw_perf_counter - performance counter hardware details
+ * struct hw_perf_counter - performance counter hardware details:
  */
 struct hw_perf_counter {
-	u64			config;
-	unsigned long		config_base;
-	unsigned long		counter_base;
-	int			nmi;
-	unsigned int		idx;
-	u64			prev_count;
-	s32			next_count;
-	u64			irq_period;
+	u64				config;
+	unsigned long			config_base;
+	unsigned long			counter_base;
+	int				nmi;
+	unsigned int			idx;
+	u64				prev_count;
+	u64				irq_period;
+	s32				next_count;
 };
 
 /*
  * Hardcoded buffer length limit for now, for IRQ-fed events:
  */
-#define PERF_DATA_BUFLEN	2048
+#define PERF_DATA_BUFLEN		2048
 
 /**
  * struct perf_data - performance counter IRQ data sampling ...
  */
 struct perf_data {
-	int			len;
-	int			rd_idx;
-	int			overrun;
-	u8			data[PERF_DATA_BUFLEN];
+	int				len;
+	int				rd_idx;
+	int				overrun;
+	u8				data[PERF_DATA_BUFLEN];
 };
 
 /**
@@ -96,7 +124,7 @@ struct perf_counter {
 #else
 	atomic_t			count32[2];
 #endif
-	struct perf_counter_event	event;
+	struct perf_counter_hw_event	hw_event;
 	struct hw_perf_counter		hw;
 
 	struct perf_counter_context	*ctx;
@@ -110,8 +138,6 @@ struct perf_counter {
 	int				oncpu;
 	int				cpu;
 
-	enum perf_record_type		record_type;
-
 	/* read() / irq related data */
 	wait_queue_head_t		waitq;
 	/* optional: for NMIs */

include/linux/syscalls.h

@@ -54,7 +54,7 @@ struct compat_stat;
 struct compat_timeval;
 struct robust_list_head;
 struct getcpu_cache;
-struct perf_counter_event;
+struct perf_counter_hw_event;
 
 #include <linux/types.h>
 #include <linux/aio_abi.h>
@@ -625,7 +625,11 @@ asmlinkage long sys_fallocate(int fd, int mode, loff_t offset, loff_t len);
 
 int kernel_execve(const char *filename, char *const argv[], char *const envp[]);
 
-asmlinkage int
-sys_perf_counter_open(struct perf_counter_event __user *uevent, u32 record_type,
-		      pid_t pid, int cpu, int masterfd);
+
+asmlinkage int sys_perf_counter_open(
+
+	struct perf_counter_hw_event	*hw_event_uptr		__user,
+	pid_t				pid,
+	int				cpu,
+	int				group_fd);
 #endif

kernel/perf_counter.c

@@ -669,7 +669,7 @@ perf_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
 {
 	struct perf_counter *counter = file->private_data;
 
-	switch (counter->record_type) {
+	switch (counter->hw_event.record_type) {
 	case PERF_RECORD_SIMPLE:
 		return perf_read_hw(counter, buf, count);
 
@@ -707,7 +707,7 @@ static const struct file_operations perf_fops = {
  * Allocate and initialize a counter structure
  */
 static struct perf_counter *
-perf_counter_alloc(struct perf_counter_event *event, int cpu, u32 record_type)
+perf_counter_alloc(struct perf_counter_hw_event *hw_event, int cpu)
 {
 	struct perf_counter *counter = kzalloc(sizeof(*counter), GFP_KERNEL);
 
@@ -718,31 +718,37 @@ perf_counter_alloc(struct perf_counter_event *event, int cpu, u32 record_type)
 	INIT_LIST_HEAD(&counter->list);
 	init_waitqueue_head(&counter->waitq);
 
-	counter->irqdata	= &counter->data[0];
-	counter->usrdata	= &counter->data[1];
-	counter->cpu		= cpu;
-	counter->record_type	= record_type;
-	counter->event		= *event;
-	counter->wakeup_pending = 0;
+	counter->irqdata		= &counter->data[0];
+	counter->usrdata		= &counter->data[1];
+	counter->cpu			= cpu;
+	counter->hw_event		= *hw_event;
+	counter->wakeup_pending		= 0;
 
 	return counter;
 }
 
 /**
- * sys_perf_task_open - open a performance counter associate it to a task
- * @hw_event_type:	event type for monitoring/sampling...
+ * sys_perf_task_open - open a performance counter, associate it to a task/cpu
+ *
+ * @hw_event_uptr:	event type attributes for monitoring/sampling
  * @pid:		target pid
+ * @cpu:		target cpu
+ * @group_fd:		group leader counter fd
  */
-asmlinkage int
-sys_perf_counter_open(struct perf_counter_event __user *uevent, u32 record_type,
-		      pid_t pid, int cpu, int masterfd)
+asmlinkage int sys_perf_counter_open(
+
+	struct perf_counter_hw_event	*hw_event_uptr		__user,
+	pid_t				pid,
+	int				cpu,
+	int				group_fd)
+
 {
 	struct perf_counter_context *ctx;
-	struct perf_counter_event event;
+	struct perf_counter_hw_event hw_event;
 	struct perf_counter *counter;
 	int ret;
 
-	if (copy_from_user(&event, uevent, sizeof(event)) != 0)
+	if (copy_from_user(&hw_event, hw_event_uptr, sizeof(hw_event)) != 0)
 		return -EFAULT;
 
 	ctx = find_get_context(pid, cpu);
@@ -750,7 +756,7 @@ sys_perf_counter_open(struct perf_counter_event __user *uevent, u32 record_type,
 		return PTR_ERR(ctx);
 
 	ret = -ENOMEM;
-	counter = perf_counter_alloc(&event, cpu, record_type);
+	counter = perf_counter_alloc(&hw_event, cpu);
 	if (!counter)
 		goto err_put_context;