Merge branch 'perf-fixes-for-linus' of...

Merge branch 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (57 commits)
  x86, perf events: Check if we have APIC enabled
  perf_event: Fix variable initialization in other codepaths
  perf kmem: Fix unused argument build warning
  perf symbols: perf_header__read_build_ids() offset'n'size should be u64
  perf symbols: dsos__read_build_ids() should read both user and kernel buildids
  perf tools: Align long options which have no short forms
  perf kmem: Show usage if no option is specified
  sched: Mark sched_clock() as notrace
  perf sched: Add max delay time snapshot
  perf tools: Correct size given to memset
  perf_event: Fix perf_swevent_hrtimer() variable initialization
  perf sched: Fix for getting task's execution time
  tracing/kprobes: Fix field creation's bad error handling
  perf_event: Cleanup for cpu_clock_perf_event_update()
  perf_event: Allocate children's perf_event_ctxp at the right time
  perf_event: Clean up __perf_event_init_context()
  hw-breakpoints: Modify breakpoints without unregistering them
  perf probe: Update perf-probe document
  perf probe: Support --del option
  trace-kprobe: Support delete probe syntax
  ...

arch/x86/Kconfig.debug

@@ -187,8 +187,8 @@ config HAVE_MMIOTRACE_SUPPORT
 	def_bool y
 
 config X86_DECODER_SELFTEST
-     bool "x86 instruction decoder selftest"
-     depends on DEBUG_KERNEL
+	bool "x86 instruction decoder selftest"
+	depends on DEBUG_KERNEL && KPROBES
 	---help---
 	 Perform x86 instruction decoder selftests at build time.
 	 This option is useful for checking the sanity of x86 instruction

arch/x86/kernel/cpu/perf_event.c

@@ -1632,6 +1632,7 @@ static void intel_pmu_drain_bts_buffer(struct cpu_hw_events *cpuc)
 
 	data.period	= event->hw.last_period;
 	data.addr	= 0;
+	data.raw	= NULL;
 	regs.ip		= 0;
 
 	/*
@@ -1749,6 +1750,7 @@ static int p6_pmu_handle_irq(struct pt_regs *regs)
 	u64 val;
 
 	data.addr = 0;
+	data.raw = NULL;
 
 	cpuc = &__get_cpu_var(cpu_hw_events);
 
@@ -1794,6 +1796,7 @@ static int intel_pmu_handle_irq(struct pt_regs *regs)
 	u64 ack, status;
 
 	data.addr = 0;
+	data.raw = NULL;
 
 	cpuc = &__get_cpu_var(cpu_hw_events);
 
@@ -1857,6 +1860,7 @@ static int amd_pmu_handle_irq(struct pt_regs *regs)
 	u64 val;
 
 	data.addr = 0;
+	data.raw = NULL;
 
 	cpuc = &__get_cpu_var(cpu_hw_events);
 
@@ -2062,12 +2066,6 @@ static __init int p6_pmu_init(void)
 
 	x86_pmu = p6_pmu;
 
-	if (!cpu_has_apic) {
-		pr_info("no APIC, boot with the \"lapic\" boot parameter to force-enable it.\n");
-		pr_info("no hardware sampling interrupt available.\n");
-		x86_pmu.apic = 0;
-	}
-
 	return 0;
 }
 
@@ -2159,6 +2157,16 @@ static __init int amd_pmu_init(void)
 	return 0;
 }
 
+static void __init pmu_check_apic(void)
+{
+	if (cpu_has_apic)
+		return;
+
+	x86_pmu.apic = 0;
+	pr_info("no APIC, boot with the \"lapic\" boot parameter to force-enable it.\n");
+	pr_info("no hardware sampling interrupt available.\n");
+}
+
 void __init init_hw_perf_events(void)
 {
 	int err;
@@ -2180,6 +2188,8 @@ void __init init_hw_perf_events(void)
 		return;
 	}
 
+	pmu_check_apic();
+
 	pr_cont("%s PMU driver.\n", x86_pmu.name);
 
 	if (x86_pmu.num_events > X86_PMC_MAX_GENERIC) {
@@ -2287,7 +2297,7 @@ void callchain_store(struct perf_callchain_entry *entry, u64 ip)
 
 static DEFINE_PER_CPU(struct perf_callchain_entry, pmc_irq_entry);
 static DEFINE_PER_CPU(struct perf_callchain_entry, pmc_nmi_entry);
-static DEFINE_PER_CPU(int, in_nmi_frame);
+static DEFINE_PER_CPU(int, in_ignored_frame);
 
 
 static void
@@ -2303,8 +2313,9 @@ static void backtrace_warning(void *data, char *msg)
 
 static int backtrace_stack(void *data, char *name)
 {
-	per_cpu(in_nmi_frame, smp_processor_id()) =
-			x86_is_stack_id(NMI_STACK, name);
+	per_cpu(in_ignored_frame, smp_processor_id()) =
+			x86_is_stack_id(NMI_STACK, name) ||
+			x86_is_stack_id(DEBUG_STACK, name);
 
 	return 0;
 }
@@ -2313,7 +2324,7 @@ static void backtrace_address(void *data, unsigned long addr, int reliable)
 {
 	struct perf_callchain_entry *entry = data;
 
-	if (per_cpu(in_nmi_frame, smp_processor_id()))
+	if (per_cpu(in_ignored_frame, smp_processor_id()))
 		return;
 
 	if (reliable)

arch/x86/kernel/dumpstack_64.c

@@ -103,6 +103,35 @@ static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
 	return NULL;
 }
 
+static inline int
+in_irq_stack(unsigned long *stack, unsigned long *irq_stack,
+	     unsigned long *irq_stack_end)
+{
+	return (stack >= irq_stack && stack < irq_stack_end);
+}
+
+/*
+ * We are returning from the irq stack and go to the previous one.
+ * If the previous stack is also in the irq stack, then bp in the first
+ * frame of the irq stack points to the previous, interrupted one.
+ * Otherwise we have another level of indirection: We first save
+ * the bp of the previous stack, then we switch the stack to the irq one
+ * and save a new bp that links to the previous one.
+ * (See save_args())
+ */
+static inline unsigned long
+fixup_bp_irq_link(unsigned long bp, unsigned long *stack,
+		  unsigned long *irq_stack, unsigned long *irq_stack_end)
+{
+#ifdef CONFIG_FRAME_POINTER
+	struct stack_frame *frame = (struct stack_frame *)bp;
+
+	if (!in_irq_stack(stack, irq_stack, irq_stack_end))
+		return (unsigned long)frame->next_frame;
+#endif
+	return bp;
+}
+
 /*
  * x86-64 can have up to three kernel stacks:
  * process stack
@@ -175,7 +204,7 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
 			irq_stack = irq_stack_end -
 				(IRQ_STACK_SIZE - 64) / sizeof(*irq_stack);
 
-			if (stack >= irq_stack && stack < irq_stack_end) {
+			if (in_irq_stack(stack, irq_stack, irq_stack_end)) {
 				if (ops->stack(data, "IRQ") < 0)
 					break;
 				bp = print_context_stack(tinfo, stack, bp,
@@ -186,6 +215,8 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
 				 * pointer (index -1 to end) in the IRQ stack:
 				 */
 				stack = (unsigned long *) (irq_stack_end[-1]);
+				bp = fixup_bp_irq_link(bp, stack, irq_stack,
+						       irq_stack_end);
 				irq_stack_end = NULL;
 				ops->stack(data, "EOI");
 				continue;

arch/x86/kernel/entry_64.S

@@ -1076,10 +1076,10 @@ ENTRY(\sym)
 	TRACE_IRQS_OFF
 	movq %rsp,%rdi		/* pt_regs pointer */
 	xorl %esi,%esi		/* no error code */
-	PER_CPU(init_tss, %rbp)
-	subq $EXCEPTION_STKSZ, TSS_ist + (\ist - 1) * 8(%rbp)
+	PER_CPU(init_tss, %r12)
+	subq $EXCEPTION_STKSZ, TSS_ist + (\ist - 1) * 8(%r12)
 	call \do_sym
-	addq $EXCEPTION_STKSZ, TSS_ist + (\ist - 1) * 8(%rbp)
+	addq $EXCEPTION_STKSZ, TSS_ist + (\ist - 1) * 8(%r12)
 	jmp paranoid_exit	/* %ebx: no swapgs flag */
 	CFI_ENDPROC
 END(\sym)

arch/x86/kernel/hw_breakpoint.c

@@ -362,8 +362,7 @@ int arch_validate_hwbkpt_settings(struct perf_event *bp,
 		return ret;
 	}
 
-	if (bp->callback)
-		ret = arch_store_info(bp);
+	ret = arch_store_info(bp);
 
 	if (ret < 0)
 		return ret;
@@ -519,7 +518,7 @@ static int __kprobes hw_breakpoint_handler(struct die_args *args)
 			break;
 		}
 
-		(bp->callback)(bp, args->regs);
+		perf_bp_event(bp, args->regs);
 
 		rcu_read_unlock();
 	}

arch/x86/kernel/ptrace.c

@@ -555,7 +555,9 @@ static int genregs_set(struct task_struct *target,
 	return ret;
 }
 
-static void ptrace_triggered(struct perf_event *bp, void *data)
+static void ptrace_triggered(struct perf_event *bp, int nmi,
+			     struct perf_sample_data *data,
+			     struct pt_regs *regs)
 {
 	int i;
 	struct thread_struct *thread = &(current->thread);
@@ -593,13 +595,13 @@ static unsigned long ptrace_get_dr7(struct perf_event *bp[])
 	return dr7;
 }
 
-static struct perf_event *
+static int
 ptrace_modify_breakpoint(struct perf_event *bp, int len, int type,
 			 struct task_struct *tsk, int disabled)
 {
 	int err;
 	int gen_len, gen_type;
-	DEFINE_BREAKPOINT_ATTR(attr);
+	struct perf_event_attr attr;
 
 	/*
 	 * We shoud have at least an inactive breakpoint at this
@@ -607,18 +609,18 @@ ptrace_modify_breakpoint(struct perf_event *bp, int len, int type,
 	 * written the address register first
 	 */
 	if (!bp)
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
 	err = arch_bp_generic_fields(len, type, &gen_len, &gen_type);
 	if (err)
-		return ERR_PTR(err);
+		return err;
 
 	attr = bp->attr;
 	attr.bp_len = gen_len;
 	attr.bp_type = gen_type;
 	attr.disabled = disabled;
 
-	return modify_user_hw_breakpoint(bp, &attr, bp->callback, tsk);
+	return modify_user_hw_breakpoint(bp, &attr);
 }
 
 /*
@@ -656,28 +658,17 @@ static int ptrace_write_dr7(struct task_struct *tsk, unsigned long data)
 				if (!second_pass)
 					continue;
 
-				thread->ptrace_bps[i] = NULL;
-				bp = ptrace_modify_breakpoint(bp, len, type,
+				rc = ptrace_modify_breakpoint(bp, len, type,
 							      tsk, 1);
-				if (IS_ERR(bp)) {
-					rc = PTR_ERR(bp);
-					thread->ptrace_bps[i] = NULL;
+				if (rc)
 					break;
-				}
-				thread->ptrace_bps[i] = bp;
 			}
 			continue;
 		}
 
-		bp = ptrace_modify_breakpoint(bp, len, type, tsk, 0);
-
-		/* Incorrect bp, or we have a bug in bp API */
-		if (IS_ERR(bp)) {
-			rc = PTR_ERR(bp);
-			thread->ptrace_bps[i] = NULL;
+		rc = ptrace_modify_breakpoint(bp, len, type, tsk, 0);
+		if (rc)
 			break;
-		}
-		thread->ptrace_bps[i] = bp;
 	}
 	/*
 	 * Make a second pass to free the remaining unused breakpoints
@@ -721,9 +712,10 @@ static int ptrace_set_breakpoint_addr(struct task_struct *tsk, int nr,
 {
 	struct perf_event *bp;
 	struct thread_struct *t = &tsk->thread;
-	DEFINE_BREAKPOINT_ATTR(attr);
+	struct perf_event_attr attr;
 
 	if (!t->ptrace_bps[nr]) {
+		hw_breakpoint_init(&attr);
 		/*
 		 * Put stub len and type to register (reserve) an inactive but
 		 * correct bp
@@ -734,26 +726,32 @@ static int ptrace_set_breakpoint_addr(struct task_struct *tsk, int nr,
 		attr.disabled = 1;
 
 		bp = register_user_hw_breakpoint(&attr, ptrace_triggered, tsk);
+
+		/*
+		 * CHECKME: the previous code returned -EIO if the addr wasn't
+		 * a valid task virtual addr. The new one will return -EINVAL in
+		 *  this case.
+		 * -EINVAL may be what we want for in-kernel breakpoints users,
+		 * but -EIO looks better for ptrace, since we refuse a register
+		 * writing for the user. And anyway this is the previous
+		 * behaviour.
+		 */
+		if (IS_ERR(bp))
+			return PTR_ERR(bp);
+
+		t->ptrace_bps[nr] = bp;
 	} else {
+		int err;
+
 		bp = t->ptrace_bps[nr];
-		t->ptrace_bps[nr] = NULL;
 
 		attr = bp->attr;
 		attr.bp_addr = addr;
-		bp = modify_user_hw_breakpoint(bp, &attr, bp->callback, tsk);
+		err = modify_user_hw_breakpoint(bp, &attr);
+		if (err)
+			return err;
 	}
-	/*
-	 * CHECKME: the previous code returned -EIO if the addr wasn't a
-	 * valid task virtual addr. The new one will return -EINVAL in this
-	 * case.
-	 * -EINVAL may be what we want for in-kernel breakpoints users, but
-	 * -EIO looks better for ptrace, since we refuse a register writing
-	 * for the user. And anyway this is the previous behaviour.
-	 */
-	if (IS_ERR(bp))
-		return PTR_ERR(bp);
 
-	t->ptrace_bps[nr] = bp;
 
 	return 0;
 }

arch/x86/lib/Makefile

@@ -5,7 +5,7 @@
 inat_tables_script = $(srctree)/arch/x86/tools/gen-insn-attr-x86.awk
 inat_tables_maps = $(srctree)/arch/x86/lib/x86-opcode-map.txt
 quiet_cmd_inat_tables = GEN     $@
-      cmd_inat_tables = $(AWK) -f $(inat_tables_script) $(inat_tables_maps) > $@
+      cmd_inat_tables = $(AWK) -f $(inat_tables_script) $(inat_tables_maps) > $@ || rm -f $@
 
 $(obj)/inat-tables.c: $(inat_tables_script) $(inat_tables_maps)
 	$(call cmd,inat_tables)
@@ -20,7 +20,7 @@ lib-y := delay.o
 lib-y += thunk_$(BITS).o
 lib-y += usercopy_$(BITS).o getuser.o putuser.o
 lib-y += memcpy_$(BITS).o
-lib-y += insn.o inat.o
+lib-$(CONFIG_KPROBES) += insn.o inat.o
 
 obj-y += msr-reg.o msr-reg-export.o
 

arch/x86/tools/test_get_len.c

@@ -113,7 +113,7 @@ int main(int argc, char **argv)
 	char line[BUFSIZE], sym[BUFSIZE] = "<unknown>";
 	unsigned char insn_buf[16];
 	struct insn insn;
-	int insns = 0, c;
+	int insns = 0;
 	int warnings = 0;
 
 	parse_args(argc, argv);

include/linux/hw_breakpoint.h

@@ -20,19 +20,18 @@ enum {
 
 #ifdef CONFIG_HAVE_HW_BREAKPOINT
 
-/* As it's for in-kernel or ptrace use, we want it to be pinned */
-#define DEFINE_BREAKPOINT_ATTR(name)	\
-struct perf_event_attr name = {		\
-	.type = PERF_TYPE_BREAKPOINT,	\
-	.size = sizeof(name),		\
-	.pinned = 1,			\
-};
-
 static inline void hw_breakpoint_init(struct perf_event_attr *attr)
 {
+	memset(attr, 0, sizeof(*attr));
+
 	attr->type = PERF_TYPE_BREAKPOINT;
 	attr->size = sizeof(*attr);
+	/*
+	 * As it's for in-kernel or ptrace use, we want it to be pinned
+	 * and to call its callback every hits.
+	 */
 	attr->pinned = 1;
+	attr->sample_period = 1;
 }
 
 static inline unsigned long hw_breakpoint_addr(struct perf_event *bp)
@@ -52,27 +51,24 @@ static inline int hw_breakpoint_len(struct perf_event *bp)
 
 extern struct perf_event *
 register_user_hw_breakpoint(struct perf_event_attr *attr,
-			    perf_callback_t triggered,
+			    perf_overflow_handler_t triggered,
 			    struct task_struct *tsk);
 
 /* FIXME: only change from the attr, and don't unregister */
-extern struct perf_event *
-modify_user_hw_breakpoint(struct perf_event *bp,
-			  struct perf_event_attr *attr,
-			  perf_callback_t triggered,
-			  struct task_struct *tsk);
+extern int
+modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr);
 
 /*
  * Kernel breakpoints are not associated with any particular thread.
  */
 extern struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,
-				perf_callback_t triggered,
+				perf_overflow_handler_t	triggered,
 				int cpu);
 
 extern struct perf_event **
 register_wide_hw_breakpoint(struct perf_event_attr *attr,
-			    perf_callback_t triggered);
+			    perf_overflow_handler_t triggered);
 
 extern int register_perf_hw_breakpoint(struct perf_event *bp);
 extern int __register_perf_hw_breakpoint(struct perf_event *bp);
@@ -93,20 +89,18 @@ static inline struct arch_hw_breakpoint *counter_arch_bp(struct perf_event *bp)
 
 static inline struct perf_event *
 register_user_hw_breakpoint(struct perf_event_attr *attr,
-			    perf_callback_t triggered,
+			    perf_overflow_handler_t triggered,
 			    struct task_struct *tsk)	{ return NULL; }
-static inline struct perf_event *
+static inline int
 modify_user_hw_breakpoint(struct perf_event *bp,
-			  struct perf_event_attr *attr,
-			  perf_callback_t triggered,
-			  struct task_struct *tsk)	{ return NULL; }
+			  struct perf_event_attr *attr)	{ return NULL; }
 static inline struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,
-				perf_callback_t triggered,
+				perf_overflow_handler_t	 triggered,
 				int cpu)		{ return NULL; }
 static inline struct perf_event **
 register_wide_hw_breakpoint(struct perf_event_attr *attr,
-			    perf_callback_t triggered)	{ return NULL; }
+			    perf_overflow_handler_t triggered)	{ return NULL; }
 static inline int
 register_perf_hw_breakpoint(struct perf_event *bp)	{ return -ENOSYS; }
 static inline int

include/linux/perf_event.h

@@ -18,10 +18,6 @@
 #include <linux/ioctl.h>
 #include <asm/byteorder.h>
 
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
-#include <asm/hw_breakpoint.h>
-#endif
-
 /*
  * User-space ABI bits:
  */
@@ -215,12 +211,12 @@ struct perf_event_attr {
 		__u32		wakeup_watermark; /* bytes before wakeup   */
 	};
 
-	union {
-		struct { /* Hardware breakpoint info */
-			__u64		bp_addr;
-			__u32		bp_type;
-			__u32		bp_len;
-		};
+	struct { /* Hardware breakpoint info */
+		__u64		bp_addr;
+		__u32		bp_type;
+		__u32		bp_len;
+		__u64		__bp_reserved_1;
+		__u64		__bp_reserved_2;
 	};
 
 	__u32			__reserved_2;
@@ -451,6 +447,10 @@ enum perf_callchain_context {
 # include <asm/perf_event.h>
 #endif
 
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#include <asm/hw_breakpoint.h>
+#endif
+
 #include <linux/list.h>
 #include <linux/mutex.h>
 #include <linux/rculist.h>
@@ -565,10 +565,12 @@ struct perf_pending_entry {
 	void (*func)(struct perf_pending_entry *);
 };
 
-typedef void (*perf_callback_t)(struct perf_event *, void *);
-
 struct perf_sample_data;
 
+typedef void (*perf_overflow_handler_t)(struct perf_event *, int,
+					struct perf_sample_data *,
+					struct pt_regs *regs);
+
 /**
  * struct perf_event - performance event kernel representation:
  */
@@ -660,18 +662,12 @@ struct perf_event {
 	struct pid_namespace		*ns;
 	u64				id;
 
-	void (*overflow_handler)(struct perf_event *event,
-			int nmi, struct perf_sample_data *data,
-			struct pt_regs *regs);
+	perf_overflow_handler_t		overflow_handler;
 
 #ifdef CONFIG_EVENT_PROFILE
 	struct event_filter		*filter;
 #endif
 
-	perf_callback_t			callback;
-
-	perf_callback_t			event_callback;
-
 #endif /* CONFIG_PERF_EVENTS */
 };
 
@@ -781,7 +777,7 @@ extern struct perf_event *
 perf_event_create_kernel_counter(struct perf_event_attr *attr,
 				int cpu,
 				pid_t pid,
-				perf_callback_t callback);
+				perf_overflow_handler_t callback);
 extern u64 perf_event_read_value(struct perf_event *event,
 				 u64 *enabled, u64 *running);
 
@@ -876,6 +872,8 @@ extern void perf_output_copy(struct perf_output_handle *handle,
 			     const void *buf, unsigned int len);
 extern int perf_swevent_get_recursion_context(void);
 extern void perf_swevent_put_recursion_context(int rctx);
+extern void perf_event_enable(struct perf_event *event);
+extern void perf_event_disable(struct perf_event *event);
 #else
 static inline void
 perf_event_task_sched_in(struct task_struct *task, int cpu)		{ }
@@ -906,7 +904,8 @@ static inline void perf_event_fork(struct task_struct *tsk)		{ }
 static inline void perf_event_init(void)				{ }
 static inline int  perf_swevent_get_recursion_context(void)  { return -1; }
 static inline void perf_swevent_put_recursion_context(int rctx)		{ }
-
+static inline void perf_event_enable(struct perf_event *event)		{ }
+static inline void perf_event_disable(struct perf_event *event)		{ }
 #endif
 
 #define perf_output_put(handle, x) \

include/linux/sched.h

@@ -1840,7 +1840,8 @@ static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask)
 extern int sched_clock_stable;
 #endif
 
-extern unsigned long long sched_clock(void);
+/* ftrace calls sched_clock() directly */
+extern unsigned long long notrace sched_clock(void);
 
 extern void sched_clock_init(void);
 extern u64 sched_clock_cpu(int cpu);

kernel/hw_breakpoint.c

@@ -52,7 +52,7 @@
 static DEFINE_PER_CPU(unsigned int, nr_cpu_bp_pinned);
 
 /* Number of pinned task breakpoints in a cpu */
-static DEFINE_PER_CPU(unsigned int, task_bp_pinned[HBP_NUM]);
+static DEFINE_PER_CPU(unsigned int, nr_task_bp_pinned[HBP_NUM]);
 
 /* Number of non-pinned cpu/task breakpoints in a cpu */
 static DEFINE_PER_CPU(unsigned int, nr_bp_flexible);
@@ -73,7 +73,7 @@ static DEFINE_MUTEX(nr_bp_mutex);
 static unsigned int max_task_bp_pinned(int cpu)
 {
 	int i;
-	unsigned int *tsk_pinned = per_cpu(task_bp_pinned, cpu);
+	unsigned int *tsk_pinned = per_cpu(nr_task_bp_pinned, cpu);
 
 	for (i = HBP_NUM -1; i >= 0; i--) {
 		if (tsk_pinned[i] > 0)
@@ -83,15 +83,51 @@ static unsigned int max_task_bp_pinned(int cpu)
 	return 0;
 }
 
+static int task_bp_pinned(struct task_struct *tsk)
+{
+	struct perf_event_context *ctx = tsk->perf_event_ctxp;
+	struct list_head *list;
+	struct perf_event *bp;
+	unsigned long flags;
+	int count = 0;
+
+	if (WARN_ONCE(!ctx, "No perf context for this task"))
+		return 0;
+
+	list = &ctx->event_list;
+
+	spin_lock_irqsave(&ctx->lock, flags);
+
+	/*
+	 * The current breakpoint counter is not included in the list
+	 * at the open() callback time
+	 */
+	list_for_each_entry(bp, list, event_entry) {
+		if (bp->attr.type == PERF_TYPE_BREAKPOINT)
+			count++;
+	}
+
+	spin_unlock_irqrestore(&ctx->lock, flags);
+
+	return count;
+}
+
 /*
  * Report the number of pinned/un-pinned breakpoints we have in
  * a given cpu (cpu > -1) or in all of them (cpu = -1).
  */
-static void fetch_bp_busy_slots(struct bp_busy_slots *slots, int cpu)
+static void
+fetch_bp_busy_slots(struct bp_busy_slots *slots, struct perf_event *bp)
 {
+	int cpu = bp->cpu;
+	struct task_struct *tsk = bp->ctx->task;
+
 	if (cpu >= 0) {
 		slots->pinned = per_cpu(nr_cpu_bp_pinned, cpu);
-		slots->pinned += max_task_bp_pinned(cpu);
+		if (!tsk)
+			slots->pinned += max_task_bp_pinned(cpu);
+		else
+			slots->pinned += task_bp_pinned(tsk);
 		slots->flexible = per_cpu(nr_bp_flexible, cpu);
 
 		return;
@@ -101,7 +137,10 @@ static void fetch_bp_busy_slots(struct bp_busy_slots *slots, int cpu)
 		unsigned int nr;
 
 		nr = per_cpu(nr_cpu_bp_pinned, cpu);
-		nr += max_task_bp_pinned(cpu);
+		if (!tsk)
+			nr += max_task_bp_pinned(cpu);
+		else
+			nr += task_bp_pinned(tsk);
 
 		if (nr > slots->pinned)
 			slots->pinned = nr;
@@ -118,35 +157,12 @@ static void fetch_bp_busy_slots(struct bp_busy_slots *slots, int cpu)
  */
 static void toggle_bp_task_slot(struct task_struct *tsk, int cpu, bool enable)
 {
-	int count = 0;
-	struct perf_event *bp;
-	struct perf_event_context *ctx = tsk->perf_event_ctxp;
 	unsigned int *tsk_pinned;
-	struct list_head *list;
-	unsigned long flags;
-
-	if (WARN_ONCE(!ctx, "No perf context for this task"))
-		return;
-
-	list = &ctx->event_list;
-
-	spin_lock_irqsave(&ctx->lock, flags);
-
-	/*
-	 * The current breakpoint counter is not included in the list
-	 * at the open() callback time
-	 */
-	list_for_each_entry(bp, list, event_entry) {
-		if (bp->attr.type == PERF_TYPE_BREAKPOINT)
-			count++;
-	}
+	int count = 0;
 
-	spin_unlock_irqrestore(&ctx->lock, flags);
+	count = task_bp_pinned(tsk);
 
-	if (WARN_ONCE(count < 0, "No breakpoint counter found in the counter list"))
-		return;
-
-	tsk_pinned = per_cpu(task_bp_pinned, cpu);
+	tsk_pinned = per_cpu(nr_task_bp_pinned, cpu);
 	if (enable) {
 		tsk_pinned[count]++;
 		if (count > 0)
@@ -193,7 +209,7 @@ static void toggle_bp_slot(struct perf_event *bp, bool enable)
  *   - If attached to a single cpu, check:
  *
  *       (per_cpu(nr_bp_flexible, cpu) || (per_cpu(nr_cpu_bp_pinned, cpu)
- *           + max(per_cpu(task_bp_pinned, cpu)))) < HBP_NUM
+ *           + max(per_cpu(nr_task_bp_pinned, cpu)))) < HBP_NUM
  *
  *       -> If there are already non-pinned counters in this cpu, it means
  *          there is already a free slot for them.
@@ -204,7 +220,7 @@ static void toggle_bp_slot(struct perf_event *bp, bool enable)
  *   - If attached to every cpus, check:
  *
  *       (per_cpu(nr_bp_flexible, *) || (max(per_cpu(nr_cpu_bp_pinned, *))
- *           + max(per_cpu(task_bp_pinned, *)))) < HBP_NUM
+ *           + max(per_cpu(nr_task_bp_pinned, *)))) < HBP_NUM
  *
  *       -> This is roughly the same, except we check the number of per cpu
  *          bp for every cpu and we keep the max one. Same for the per tasks
@@ -216,7 +232,7 @@ static void toggle_bp_slot(struct perf_event *bp, bool enable)
  *   - If attached to a single cpu, check:
  *
  *       ((per_cpu(nr_bp_flexible, cpu) > 1) + per_cpu(nr_cpu_bp_pinned, cpu)
- *            + max(per_cpu(task_bp_pinned, cpu))) < HBP_NUM
+ *            + max(per_cpu(nr_task_bp_pinned, cpu))) < HBP_NUM
  *
  *       -> Same checks as before. But now the nr_bp_flexible, if any, must keep
  *          one register at least (or they will never be fed).
@@ -224,7 +240,7 @@ static void toggle_bp_slot(struct perf_event *bp, bool enable)
  *   - If attached to every cpus, check:
  *
  *       ((per_cpu(nr_bp_flexible, *) > 1) + max(per_cpu(nr_cpu_bp_pinned, *))
- *            + max(per_cpu(task_bp_pinned, *))) < HBP_NUM
+ *            + max(per_cpu(nr_task_bp_pinned, *))) < HBP_NUM
  */
 int reserve_bp_slot(struct perf_event *bp)
 {
@@ -233,7 +249,7 @@ int reserve_bp_slot(struct perf_event *bp)
 
 	mutex_lock(&nr_bp_mutex);
 
-	fetch_bp_busy_slots(&slots, bp->cpu);
+	fetch_bp_busy_slots(&slots, bp);
 
 	/* Flexible counters need to keep at least one slot */
 	if (slots.pinned + (!!slots.flexible) == HBP_NUM) {
@@ -259,7 +275,7 @@ void release_bp_slot(struct perf_event *bp)
 }
 
 
-int __register_perf_hw_breakpoint(struct perf_event *bp)
+int register_perf_hw_breakpoint(struct perf_event *bp)
 {
 	int ret;
 
@@ -276,19 +292,12 @@ int __register_perf_hw_breakpoint(struct perf_event *bp)
 	 * This is a quick hack that will be removed soon, once we remove
 	 * the tmp breakpoints from ptrace
 	 */
-	if (!bp->attr.disabled || bp->callback == perf_bp_event)
+	if (!bp->attr.disabled || !bp->overflow_handler)
 		ret = arch_validate_hwbkpt_settings(bp, bp->ctx->task);
 
 	return ret;
 }
 
-int register_perf_hw_breakpoint(struct perf_event *bp)
-{
-	bp->callback = perf_bp_event;
-
-	return __register_perf_hw_breakpoint(bp);
-}
-
 /**
  * register_user_hw_breakpoint - register a hardware breakpoint for user space
  * @attr: breakpoint attributes
@@ -297,7 +306,7 @@ int register_perf_hw_breakpoint(struct perf_event *bp)
  */
 struct perf_event *
 register_user_hw_breakpoint(struct perf_event_attr *attr,
-			    perf_callback_t triggered,
+			    perf_overflow_handler_t triggered,
 			    struct task_struct *tsk)
 {
 	return perf_event_create_kernel_counter(attr, -1, tsk->pid, triggered);
@@ -311,19 +320,40 @@ EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
  * @triggered: callback to trigger when we hit the breakpoint
  * @tsk: pointer to 'task_struct' of the process to which the address belongs
  */
-struct perf_event *
-modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr,
-			  perf_callback_t triggered,
-			  struct task_struct *tsk)
+int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
 {
-	/*
-	 * FIXME: do it without unregistering
-	 * - We don't want to lose our slot
-	 * - If the new bp is incorrect, don't lose the older one
-	 */
-	unregister_hw_breakpoint(bp);
+	u64 old_addr = bp->attr.bp_addr;
+	int old_type = bp->attr.bp_type;
+	int old_len = bp->attr.bp_len;
+	int err = 0;
 
-	return perf_event_create_kernel_counter(attr, -1, tsk->pid, triggered);
+	perf_event_disable(bp);
+
+	bp->attr.bp_addr = attr->bp_addr;
+	bp->attr.bp_type = attr->bp_type;
+	bp->attr.bp_len = attr->bp_len;
+
+	if (attr->disabled)
+		goto end;
+
+	err = arch_validate_hwbkpt_settings(bp, bp->ctx->task);
+	if (!err)
+		perf_event_enable(bp);
+
+	if (err) {
+		bp->attr.bp_addr = old_addr;
+		bp->attr.bp_type = old_type;
+		bp->attr.bp_len = old_len;
+		if (!bp->attr.disabled)
+			perf_event_enable(bp);
+
+		return err;
+	}
+
+end:
+	bp->attr.disabled = attr->disabled;
+
+	return 0;
 }
 EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint);
 
@@ -348,7 +378,7 @@ EXPORT_SYMBOL_GPL(unregister_hw_breakpoint);
  */
 struct perf_event **
 register_wide_hw_breakpoint(struct perf_event_attr *attr,
-			    perf_callback_t triggered)
+			    perf_overflow_handler_t triggered)
 {
 	struct perf_event **cpu_events, **pevent, *bp;
 	long err;

kernel/perf_event.c

@@ -36,7 +36,7 @@
 /*
  * Each CPU has a list of per CPU events:
  */
-DEFINE_PER_CPU(struct perf_cpu_context, perf_cpu_context);
+static DEFINE_PER_CPU(struct perf_cpu_context, perf_cpu_context);
 
 int perf_max_events __read_mostly = 1;
 static int perf_reserved_percpu __read_mostly;
@@ -567,7 +567,7 @@ static void __perf_event_disable(void *info)
  * is the current context on this CPU and preemption is disabled,
  * hence we can't get into perf_event_task_sched_out for this context.
  */
-static void perf_event_disable(struct perf_event *event)
+void perf_event_disable(struct perf_event *event)
 {
 	struct perf_event_context *ctx = event->ctx;
 	struct task_struct *task = ctx->task;
@@ -971,7 +971,7 @@ static void __perf_event_enable(void *info)
  * perf_event_for_each_child or perf_event_for_each as described
  * for perf_event_disable.
  */
-static void perf_event_enable(struct perf_event *event)
+void perf_event_enable(struct perf_event *event)
 {
 	struct perf_event_context *ctx = event->ctx;
 	struct task_struct *task = ctx->task;
@@ -1579,7 +1579,6 @@ static void
 __perf_event_init_context(struct perf_event_context *ctx,
 			    struct task_struct *task)
 {
-	memset(ctx, 0, sizeof(*ctx));
 	spin_lock_init(&ctx->lock);
 	mutex_init(&ctx->mutex);
 	INIT_LIST_HEAD(&ctx->group_list);
@@ -1654,7 +1653,7 @@ static struct perf_event_context *find_get_context(pid_t pid, int cpu)
 	}
 
 	if (!ctx) {
-		ctx = kmalloc(sizeof(struct perf_event_context), GFP_KERNEL);
+		ctx = kzalloc(sizeof(struct perf_event_context), GFP_KERNEL);
 		err = -ENOMEM;
 		if (!ctx)
 			goto errout;
@@ -4011,6 +4010,7 @@ static enum hrtimer_restart perf_swevent_hrtimer(struct hrtimer *hrtimer)
 	event->pmu->read(event);
 
 	data.addr = 0;
+	data.raw = NULL;
 	data.period = event->hw.last_period;
 	regs = get_irq_regs();
 	/*
@@ -4080,8 +4080,7 @@ static void cpu_clock_perf_event_update(struct perf_event *event)
 	u64 now;
 
 	now = cpu_clock(cpu);
-	prev = atomic64_read(&event->hw.prev_count);
-	atomic64_set(&event->hw.prev_count, now);
+	prev = atomic64_xchg(&event->hw.prev_count, now);
 	atomic64_add(now - prev, &event->count);
 }
 
@@ -4286,15 +4285,8 @@ static void bp_perf_event_destroy(struct perf_event *event)
 static const struct pmu *bp_perf_event_init(struct perf_event *bp)
 {
 	int err;
-	/*
-	 * The breakpoint is already filled if we haven't created the counter
-	 * through perf syscall
-	 * FIXME: manage to get trigerred to NULL if it comes from syscalls
-	 */
-	if (!bp->callback)
-		err = register_perf_hw_breakpoint(bp);
-	else
-		err = __register_perf_hw_breakpoint(bp);
+
+	err = register_perf_hw_breakpoint(bp);
 	if (err)
 		return ERR_PTR(err);
 
@@ -4308,6 +4300,7 @@ void perf_bp_event(struct perf_event *bp, void *data)
 	struct perf_sample_data sample;
 	struct pt_regs *regs = data;
 
+	sample.raw = NULL;
 	sample.addr = bp->attr.bp_addr;
 
 	if (!perf_exclude_event(bp, regs))
@@ -4390,7 +4383,7 @@ perf_event_alloc(struct perf_event_attr *attr,
 		   struct perf_event_context *ctx,
 		   struct perf_event *group_leader,
 		   struct perf_event *parent_event,
-		   perf_callback_t callback,
+		   perf_overflow_handler_t overflow_handler,
 		   gfp_t gfpflags)
 {
 	const struct pmu *pmu;
@@ -4433,10 +4426,10 @@ perf_event_alloc(struct perf_event_attr *attr,
 
 	event->state		= PERF_EVENT_STATE_INACTIVE;
 
-	if (!callback && parent_event)
-		callback = parent_event->callback;
+	if (!overflow_handler && parent_event)
+		overflow_handler = parent_event->overflow_handler;
 	
-	event->callback	= callback;
+	event->overflow_handler	= overflow_handler;
 
 	if (attr->disabled)
 		event->state = PERF_EVENT_STATE_OFF;
@@ -4776,7 +4769,8 @@ SYSCALL_DEFINE5(perf_event_open,
  */
 struct perf_event *
 perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu,
-				 pid_t pid, perf_callback_t callback)
+				 pid_t pid,
+				 perf_overflow_handler_t overflow_handler)
 {
 	struct perf_event *event;
 	struct perf_event_context *ctx;
@@ -4793,7 +4787,7 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu,
 	}
 
 	event = perf_event_alloc(attr, cpu, ctx, NULL,
-				     NULL, callback, GFP_KERNEL);
+				 NULL, overflow_handler, GFP_KERNEL);
 	if (IS_ERR(event)) {
 		err = PTR_ERR(event);
 		goto err_put_context;
@@ -5090,7 +5084,7 @@ void perf_event_free_task(struct task_struct *task)
  */
 int perf_event_init_task(struct task_struct *child)
 {
-	struct perf_event_context *child_ctx, *parent_ctx;
+	struct perf_event_context *child_ctx = NULL, *parent_ctx;
 	struct perf_event_context *cloned_ctx;
 	struct perf_event *event;
 	struct task_struct *parent = current;
@@ -5105,20 +5099,6 @@ int perf_event_init_task(struct task_struct *child)
 	if (likely(!parent->perf_event_ctxp))
 		return 0;
 
-	/*
-	 * This is executed from the parent task context, so inherit
-	 * events that have been marked for cloning.
-	 * First allocate and initialize a context for the child.
-	 */
-
-	child_ctx = kmalloc(sizeof(struct perf_event_context), GFP_KERNEL);
-	if (!child_ctx)
-		return -ENOMEM;
-
-	__perf_event_init_context(child_ctx, child);
-	child->perf_event_ctxp = child_ctx;
-	get_task_struct(child);
-
 	/*
 	 * If the parent's context is a clone, pin it so it won't get
 	 * swapped under us.
@@ -5149,6 +5129,26 @@ int perf_event_init_task(struct task_struct *child)
 			continue;
 		}
 
+		if (!child->perf_event_ctxp) {
+			/*
+			 * This is executed from the parent task context, so
+			 * inherit events that have been marked for cloning.
+			 * First allocate and initialize a context for the
+			 * child.
+			 */
+
+			child_ctx = kzalloc(sizeof(struct perf_event_context),
+					    GFP_KERNEL);
+			if (!child_ctx) {
+				ret = -ENOMEM;
+				goto exit;
+			}
+
+			__perf_event_init_context(child_ctx, child);
+			child->perf_event_ctxp = child_ctx;
+			get_task_struct(child);
+		}
+
 		ret = inherit_group(event, parent, parent_ctx,
 					     child, child_ctx);
 		if (ret) {
@@ -5177,6 +5177,7 @@ int perf_event_init_task(struct task_struct *child)
 		get_ctx(child_ctx->parent_ctx);
 	}
 
+exit:
 	mutex_unlock(&parent_ctx->mutex);
 
 	perf_unpin_context(parent_ctx);

kernel/trace/trace_kprobe.c

@@ -606,23 +606,22 @@ static int create_trace_probe(int argc, char **argv)
 	 */
 	struct trace_probe *tp;
 	int i, ret = 0;
-	int is_return = 0;
+	int is_return = 0, is_delete = 0;
 	char *symbol = NULL, *event = NULL, *arg = NULL, *group = NULL;
 	unsigned long offset = 0;
 	void *addr = NULL;
 	char buf[MAX_EVENT_NAME_LEN];
 
-	if (argc < 2) {
-		pr_info("Probe point is not specified.\n");
-		return -EINVAL;
-	}
-
+	/* argc must be >= 1 */
 	if (argv[0][0] == 'p')
 		is_return = 0;
 	else if (argv[0][0] == 'r')
 		is_return = 1;
+	else if (argv[0][0] == '-')
+		is_delete = 1;
 	else {
-		pr_info("Probe definition must be started with 'p' or 'r'.\n");
+		pr_info("Probe definition must be started with 'p', 'r' or"
+			" '-'.\n");
 		return -EINVAL;
 	}
 
@@ -642,7 +641,29 @@ static int create_trace_probe(int argc, char **argv)
 			return -EINVAL;
 		}
 	}
+	if (!group)
+		group = KPROBE_EVENT_SYSTEM;
 
+	if (is_delete) {
+		if (!event) {
+			pr_info("Delete command needs an event name.\n");
+			return -EINVAL;
+		}
+		tp = find_probe_event(event, group);
+		if (!tp) {
+			pr_info("Event %s/%s doesn't exist.\n", group, event);
+			return -ENOENT;
+		}
+		/* delete an event */
+		unregister_trace_probe(tp);
+		free_trace_probe(tp);
+		return 0;
+	}
+
+	if (argc < 2) {
+		pr_info("Probe point is not specified.\n");
+		return -EINVAL;
+	}
 	if (isdigit(argv[1][0])) {
 		if (is_return) {
 			pr_info("Return probe point must be a symbol.\n");
@@ -671,8 +692,6 @@ static int create_trace_probe(int argc, char **argv)
 	argc -= 2; argv += 2;
 
 	/* setup a probe */
-	if (!group)
-		group = KPROBE_EVENT_SYSTEM;
 	if (!event) {
 		/* Make a new event name */
 		if (symbol)
@@ -1114,7 +1133,7 @@ static int kprobe_event_define_fields(struct ftrace_event_call *event_call)
 	struct trace_probe *tp = (struct trace_probe *)event_call->data;
 
 	ret = trace_define_common_fields(event_call);
-	if (!ret)
+	if (ret)
 		return ret;
 
 	DEFINE_FIELD(unsigned long, ip, FIELD_STRING_IP, 0);
@@ -1132,7 +1151,7 @@ static int kretprobe_event_define_fields(struct ftrace_event_call *event_call)
 	struct trace_probe *tp = (struct trace_probe *)event_call->data;
 
 	ret = trace_define_common_fields(event_call);
-	if (!ret)
+	if (ret)
 		return ret;
 
 	DEFINE_FIELD(unsigned long, func, FIELD_STRING_FUNC, 0);

kernel/trace/trace_ksym.c

@@ -79,11 +79,12 @@ void ksym_collect_stats(unsigned long hbp_hit_addr)
 }
 #endif /* CONFIG_PROFILE_KSYM_TRACER */
 
-void ksym_hbp_handler(struct perf_event *hbp, void *data)
+void ksym_hbp_handler(struct perf_event *hbp, int nmi,
+		      struct perf_sample_data *data,
+		      struct pt_regs *regs)
 {
 	struct ring_buffer_event *event;
 	struct ksym_trace_entry *entry;
-	struct pt_regs *regs = data;
 	struct ring_buffer *buffer;
 	int pc;
 

samples/hw_breakpoint/data_breakpoint.c

@@ -41,7 +41,9 @@ module_param_string(ksym, ksym_name, KSYM_NAME_LEN, S_IRUGO);
 MODULE_PARM_DESC(ksym, "Kernel symbol to monitor; this module will report any"
 			" write operations on the kernel symbol");
 
-static void sample_hbp_handler(struct perf_event *temp, void *data)
+static void sample_hbp_handler(struct perf_event *bp, int nmi,
+			       struct perf_sample_data *data,
+			       struct pt_regs *regs)
 {
 	printk(KERN_INFO "%s value is changed\n", ksym_name);
 	dump_stack();
@@ -51,8 +53,9 @@ static void sample_hbp_handler(struct perf_event *temp, void *data)
 static int __init hw_break_module_init(void)
 {
 	int ret;
-	DEFINE_BREAKPOINT_ATTR(attr);
+	struct perf_event_attr attr;
 
+	hw_breakpoint_init(&attr);
 	attr.bp_addr = kallsyms_lookup_name(ksym_name);
 	attr.bp_len = HW_BREAKPOINT_LEN_4;
 	attr.bp_type = HW_BREAKPOINT_W | HW_BREAKPOINT_R;

tools/perf/Documentation/perf-kmem.txt

@@ -8,16 +8,16 @@ perf-kmem - Tool to trace/measure kernel memory(slab) properties
 SYNOPSIS
 --------
 [verse]
-'perf kmem' {record} [<options>]
+'perf kmem' {record|stat} [<options>]
 
 DESCRIPTION
 -----------
-There's two variants of perf kmem:
+There are two variants of perf kmem:
 
   'perf kmem record <command>' to record the kmem events
   of an arbitrary workload.
 
-  'perf kmem' to report kernel memory statistics.
+  'perf kmem stat' to report kernel memory statistics.
 
 OPTIONS
 -------
@@ -25,8 +25,11 @@ OPTIONS
 --input=<file>::
 	Select the input file (default: perf.data)
 
---stat=<caller|alloc>::
-	Select per callsite or per allocation statistics
+--caller::
+	Show per-callsite statistics
+
+--alloc::
+	Show per-allocation statistics
 
 -s <key[,key2...]>::
 --sort=<key[,key2...]>::

tools/perf/Documentation/perf-probe.txt

@@ -8,10 +8,13 @@ perf-probe - Define new dynamic tracepoints
 SYNOPSIS
 --------
 [verse]
-'perf probe' [options] --add 'PROBE' [--add 'PROBE' ...]
+'perf probe' [options] --add='PROBE' [...]
 or
-'perf probe' [options] 'PROBE' ['PROBE' ...]
-
+'perf probe' [options] PROBE
+or
+'perf probe' [options] --del='[GROUP:]EVENT' [...]
+or
+'perf probe' --list
 
 DESCRIPTION
 -----------
@@ -31,8 +34,16 @@ OPTIONS
         Be more verbose (show parsed arguments, etc).
 
 -a::
---add::
-	Define a probe point (see PROBE SYNTAX for detail)
+--add=::
+	Define a probe event (see PROBE SYNTAX for detail).
+
+-d::
+--del=::
+	Delete a probe event.
+
+-l::
+--list::
+	List up current probe events.
 
 PROBE SYNTAX
 ------------

tools/perf/builtin-buildid-list.c

@@ -19,7 +19,7 @@ static char const *input_name = "perf.data";
 static int force;
 
 static const char *const buildid_list_usage[] = {
-	"perf report [<options>]",
+	"perf buildid-list [<options>]",
 	NULL
 };
 

tools/perf/builtin-kmem.c

@@ -57,11 +57,6 @@ static struct rb_root root_caller_sorted;
 static unsigned long total_requested, total_allocated;
 static unsigned long nr_allocs, nr_cross_allocs;
 
-struct raw_event_sample {
-	u32 size;
-	char data[0];
-};
-
 #define PATH_SYS_NODE	"/sys/devices/system/node"
 
 static void init_cpunode_map(void)
@@ -201,7 +196,7 @@ static void insert_caller_stat(unsigned long call_site,
 	}
 }
 
-static void process_alloc_event(struct raw_event_sample *raw,
+static void process_alloc_event(void *data,
 				struct event *event,
 				int cpu,
 				u64 timestamp __used,
@@ -214,10 +209,10 @@ static void process_alloc_event(struct raw_event_sample *raw,
 	int bytes_alloc;
 	int node1, node2;
 
-	ptr = raw_field_value(event, "ptr", raw->data);
-	call_site = raw_field_value(event, "call_site", raw->data);
-	bytes_req = raw_field_value(event, "bytes_req", raw->data);
-	bytes_alloc = raw_field_value(event, "bytes_alloc", raw->data);
+	ptr = raw_field_value(event, "ptr", data);
+	call_site = raw_field_value(event, "call_site", data);
+	bytes_req = raw_field_value(event, "bytes_req", data);
+	bytes_alloc = raw_field_value(event, "bytes_alloc", data);
 
 	insert_alloc_stat(call_site, ptr, bytes_req, bytes_alloc, cpu);
 	insert_caller_stat(call_site, bytes_req, bytes_alloc);
@@ -227,7 +222,7 @@ static void process_alloc_event(struct raw_event_sample *raw,
 
 	if (node) {
 		node1 = cpunode_map[cpu];
-		node2 = raw_field_value(event, "node", raw->data);
+		node2 = raw_field_value(event, "node", data);
 		if (node1 != node2)
 			nr_cross_allocs++;
 	}
@@ -262,7 +257,7 @@ static struct alloc_stat *search_alloc_stat(unsigned long ptr,
 	return NULL;
 }
 
-static void process_free_event(struct raw_event_sample *raw,
+static void process_free_event(void *data,
 			       struct event *event,
 			       int cpu,
 			       u64 timestamp __used,
@@ -271,7 +266,7 @@ static void process_free_event(struct raw_event_sample *raw,
 	unsigned long ptr;
 	struct alloc_stat *s_alloc, *s_caller;
 
-	ptr = raw_field_value(event, "ptr", raw->data);
+	ptr = raw_field_value(event, "ptr", data);
 
 	s_alloc = search_alloc_stat(ptr, 0, &root_alloc_stat, ptr_cmp);
 	if (!s_alloc)
@@ -289,66 +284,53 @@ static void process_free_event(struct raw_event_sample *raw,
 }
 
 static void
-process_raw_event(event_t *raw_event __used, void *more_data,
+process_raw_event(event_t *raw_event __used, void *data,
 		  int cpu, u64 timestamp, struct thread *thread)
 {
-	struct raw_event_sample *raw = more_data;
 	struct event *event;
 	int type;
 
-	type = trace_parse_common_type(raw->data);
+	type = trace_parse_common_type(data);
 	event = trace_find_event(type);
 
 	if (!strcmp(event->name, "kmalloc") ||
 	    !strcmp(event->name, "kmem_cache_alloc")) {
-		process_alloc_event(raw, event, cpu, timestamp, thread, 0);
+		process_alloc_event(data, event, cpu, timestamp, thread, 0);
 		return;
 	}
 
 	if (!strcmp(event->name, "kmalloc_node") ||
 	    !strcmp(event->name, "kmem_cache_alloc_node")) {
-		process_alloc_event(raw, event, cpu, timestamp, thread, 1);
+		process_alloc_event(data, event, cpu, timestamp, thread, 1);
 		return;
 	}
 
 	if (!strcmp(event->name, "kfree") ||
 	    !strcmp(event->name, "kmem_cache_free")) {
-		process_free_event(raw, event, cpu, timestamp, thread);
+		process_free_event(data, event, cpu, timestamp, thread);
 		return;
 	}
 }
 
 static int process_sample_event(event_t *event)
 {
-	u64 ip = event->ip.ip;
-	u64 timestamp = -1;
-	u32 cpu = -1;
-	u64 period = 1;
-	void *more_data = event->ip.__more_data;
-	struct thread *thread = threads__findnew(event->ip.pid);
+	struct sample_data data;
+	struct thread *thread;
 
-	if (sample_type & PERF_SAMPLE_TIME) {
-		timestamp = *(u64 *)more_data;
-		more_data += sizeof(u64);
-	}
-
-	if (sample_type & PERF_SAMPLE_CPU) {
-		cpu = *(u32 *)more_data;
-		more_data += sizeof(u32);
-		more_data += sizeof(u32); /* reserved */
-	}
+	memset(&data, 0, sizeof(data));
+	data.time = -1;
+	data.cpu = -1;
+	data.period = 1;
 
-	if (sample_type & PERF_SAMPLE_PERIOD) {
-		period = *(u64 *)more_data;
-		more_data += sizeof(u64);
-	}
+	event__parse_sample(event, sample_type, &data);
 
 	dump_printf("(IP, %d): %d/%d: %p period: %Ld\n",
 		event->header.misc,
-		event->ip.pid, event->ip.tid,
-		(void *)(long)ip,
-		(long long)period);
+		data.pid, data.tid,
+		(void *)(long)data.ip,
+		(long long)data.period);
 
+	thread = threads__findnew(event->ip.pid);
 	if (thread == NULL) {
 		pr_debug("problem processing %d event, skipping it.\n",
 			 event->header.type);
@@ -357,7 +339,8 @@ static int process_sample_event(event_t *event)
 
 	dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);
 
-	process_raw_event(event, more_data, cpu, timestamp, thread);
+	process_raw_event(event, data.raw_data, data.cpu,
+			  data.time, thread);
 
 	return 0;
 }
@@ -543,7 +526,7 @@ static int __cmd_kmem(void)
 }
 
 static const char * const kmem_usage[] = {
-	"perf kmem [<options>] {record}",
+	"perf kmem [<options>] {record|stat}",
 	NULL
 };
 
@@ -703,18 +686,17 @@ static int parse_sort_opt(const struct option *opt __used,
 	return 0;
 }
 
-static int parse_stat_opt(const struct option *opt __used,
-			  const char *arg, int unset __used)
+static int parse_caller_opt(const struct option *opt __used,
+			  const char *arg __used, int unset __used)
 {
-	if (!arg)
-		return -1;
+	caller_flag = (alloc_flag + 1);
+	return 0;
+}
 
-	if (strcmp(arg, "alloc") == 0)
-		alloc_flag = (caller_flag + 1);
-	else if (strcmp(arg, "caller") == 0)
-		caller_flag = (alloc_flag + 1);
-	else
-		return -1;
+static int parse_alloc_opt(const struct option *opt __used,
+			  const char *arg __used, int unset __used)
+{
+	alloc_flag = (caller_flag + 1);
 	return 0;
 }
 
@@ -739,14 +721,17 @@ static int parse_line_opt(const struct option *opt __used,
 static const struct option kmem_options[] = {
 	OPT_STRING('i', "input", &input_name, "file",
 		   "input file name"),
-	OPT_CALLBACK(0, "stat", NULL, "<alloc>|<caller>",
-		     "stat selector, Pass 'alloc' or 'caller'.",
-		     parse_stat_opt),
+	OPT_CALLBACK_NOOPT(0, "caller", NULL, NULL,
+			   "show per-callsite statistics",
+			   parse_caller_opt),
+	OPT_CALLBACK_NOOPT(0, "alloc", NULL, NULL,
+			   "show per-allocation statistics",
+			   parse_alloc_opt),
 	OPT_CALLBACK('s', "sort", NULL, "key[,key2...]",
 		     "sort by keys: ptr, call_site, bytes, hit, pingpong, frag",
 		     parse_sort_opt),
 	OPT_CALLBACK('l', "line", NULL, "num",
-		     "show n lins",
+		     "show n lines",
 		     parse_line_opt),
 	OPT_BOOLEAN(0, "raw-ip", &raw_ip, "show raw ip instead of symbol"),
 	OPT_END()
@@ -790,18 +775,22 @@ int cmd_kmem(int argc, const char **argv, const char *prefix __used)
 
 	argc = parse_options(argc, argv, kmem_options, kmem_usage, 0);
 
-	if (argc && !strncmp(argv[0], "rec", 3))
-		return __cmd_record(argc, argv);
-	else if (argc)
+	if (!argc)
 		usage_with_options(kmem_usage, kmem_options);
 
-	if (list_empty(&caller_sort))
-		setup_sorting(&caller_sort, default_sort_order);
-	if (list_empty(&alloc_sort))
-		setup_sorting(&alloc_sort, default_sort_order);
+	if (!strncmp(argv[0], "rec", 3)) {
+		return __cmd_record(argc, argv);
+	} else if (!strcmp(argv[0], "stat")) {
+		setup_cpunode_map();
+
+		if (list_empty(&caller_sort))
+			setup_sorting(&caller_sort, default_sort_order);
+		if (list_empty(&alloc_sort))
+			setup_sorting(&alloc_sort, default_sort_order);
 
-	setup_cpunode_map();
+		return __cmd_kmem();
+	}
 
-	return __cmd_kmem();
+	return 0;
 }