Merge branches 'tracing/fastboot', 'tracing/ftrace',...

Merge branches 'tracing/fastboot', 'tracing/ftrace', 'tracing/function-graph-tracer' and 'tracing/hw-branch-tracing' into tracing/core

arch/x86/include/asm/ds.h

@@ -6,13 +6,13 @@
  * precise-event based sampling (PEBS).
  *
  * It manages:
- * - per-thread and per-cpu allocation of BTS and PEBS
+ * - DS and BTS hardware configuration
  * - buffer overflow handling (to be done)
  * - buffer access
  *
- * It assumes:
- * - get_task_struct on all traced tasks
- * - current is allowed to trace tasks
+ * It does not do:
+ * - security checking (is the caller allowed to trace the task)
+ * - buffer allocation (memory accounting)
  *
  *
  * Copyright (C) 2007-2008 Intel Corporation.
@@ -31,6 +31,7 @@
 #ifdef CONFIG_X86_DS
 
 struct task_struct;
+struct ds_context;
 struct ds_tracer;
 struct bts_tracer;
 struct pebs_tracer;
@@ -38,6 +39,38 @@ struct pebs_tracer;
 typedef void (*bts_ovfl_callback_t)(struct bts_tracer *);
 typedef void (*pebs_ovfl_callback_t)(struct pebs_tracer *);
 
+
+/*
+ * A list of features plus corresponding macros to talk about them in
+ * the ds_request function's flags parameter.
+ *
+ * We use the enum to index an array of corresponding control bits;
+ * we use the macro to index a flags bit-vector.
+ */
+enum ds_feature {
+	dsf_bts = 0,
+	dsf_bts_kernel,
+#define BTS_KERNEL (1 << dsf_bts_kernel)
+	/* trace kernel-mode branches */
+
+	dsf_bts_user,
+#define BTS_USER (1 << dsf_bts_user)
+	/* trace user-mode branches */
+
+	dsf_bts_overflow,
+	dsf_bts_max,
+	dsf_pebs = dsf_bts_max,
+
+	dsf_pebs_max,
+	dsf_ctl_max = dsf_pebs_max,
+	dsf_bts_timestamps = dsf_ctl_max,
+#define BTS_TIMESTAMPS (1 << dsf_bts_timestamps)
+	/* add timestamps into BTS trace */
+
+#define BTS_USER_FLAGS (BTS_KERNEL | BTS_USER | BTS_TIMESTAMPS)
+};
+
+
 /*
  * Request BTS or PEBS
  *
@@ -58,92 +91,135 @@ typedef void (*pebs_ovfl_callback_t)(struct pebs_tracer *);
  *       NULL if cyclic buffer requested
  * th: the interrupt threshold in records from the end of the buffer;
  *     -1 if no interrupt threshold is requested.
+ * flags: a bit-mask of the above flags
  */
 extern struct bts_tracer *ds_request_bts(struct task_struct *task,
 					 void *base, size_t size,
-					 bts_ovfl_callback_t ovfl, size_t th);
+					 bts_ovfl_callback_t ovfl,
+					 size_t th, unsigned int flags);
 extern struct pebs_tracer *ds_request_pebs(struct task_struct *task,
 					   void *base, size_t size,
 					   pebs_ovfl_callback_t ovfl,
-					   size_t th);
+					   size_t th, unsigned int flags);
 
 /*
  * Release BTS or PEBS resources
- *
- * Returns 0 on success; -Eerrno otherwise
+ * Suspend and resume BTS or PEBS tracing
  *
  * tracer: the tracer handle returned from ds_request_~()
  */
-extern int ds_release_bts(struct bts_tracer *tracer);
-extern int ds_release_pebs(struct pebs_tracer *tracer);
+extern void ds_release_bts(struct bts_tracer *tracer);
+extern void ds_suspend_bts(struct bts_tracer *tracer);
+extern void ds_resume_bts(struct bts_tracer *tracer);
+extern void ds_release_pebs(struct pebs_tracer *tracer);
+extern void ds_suspend_pebs(struct pebs_tracer *tracer);
+extern void ds_resume_pebs(struct pebs_tracer *tracer);
+
 
 /*
- * Get the (array) index of the write pointer.
- * (assuming an array of BTS/PEBS records)
- *
- * Returns 0 on success; -Eerrno on error
+ * The raw DS buffer state as it is used for BTS and PEBS recording.
  *
- * tracer: the tracer handle returned from ds_request_~()
- * pos (out): will hold the result
+ * This is the low-level, arch-dependent interface for working
+ * directly on the raw trace data.
  */
-extern int ds_get_bts_index(struct bts_tracer *tracer, size_t *pos);
-extern int ds_get_pebs_index(struct pebs_tracer *tracer, size_t *pos);
+struct ds_trace {
+	/* the number of bts/pebs records */
+	size_t n;
+	/* the size of a bts/pebs record in bytes */
+	size_t size;
+	/* pointers into the raw buffer:
+	   - to the first entry */
+	void *begin;
+	/* - one beyond the last entry */
+	void *end;
+	/* - one beyond the newest entry */
+	void *top;
+	/* - the interrupt threshold */
+	void *ith;
+	/* flags given on ds_request() */
+	unsigned int flags;
+};
 
 /*
- * Get the (array) index one record beyond the end of the array.
- * (assuming an array of BTS/PEBS records)
- *
- * Returns 0 on success; -Eerrno on error
- *
- * tracer: the tracer handle returned from ds_request_~()
- * pos (out): will hold the result
+ * An arch-independent view on branch trace data.
  */
-extern int ds_get_bts_end(struct bts_tracer *tracer, size_t *pos);
-extern int ds_get_pebs_end(struct pebs_tracer *tracer, size_t *pos);
+enum bts_qualifier {
+	bts_invalid,
+#define BTS_INVALID bts_invalid
+
+	bts_branch,
+#define BTS_BRANCH bts_branch
+
+	bts_task_arrives,
+#define BTS_TASK_ARRIVES bts_task_arrives
+
+	bts_task_departs,
+#define BTS_TASK_DEPARTS bts_task_departs
+
+	bts_qual_bit_size = 4,
+	bts_qual_max = (1 << bts_qual_bit_size),
+};
+
+struct bts_struct {
+	__u64 qualifier;
+	union {
+		/* BTS_BRANCH */
+		struct {
+			__u64 from;
+			__u64 to;
+		} lbr;
+		/* BTS_TASK_ARRIVES or BTS_TASK_DEPARTS */
+		struct {
+			__u64 jiffies;
+			pid_t pid;
+		} timestamp;
+	} variant;
+};
+
 
 /*
- * Provide a pointer to the BTS/PEBS record at parameter index.
- * (assuming an array of BTS/PEBS records)
- *
- * The pointer points directly into the buffer. The user is
- * responsible for copying the record.
- *
- * Returns the size of a single record on success; -Eerrno on error
+ * The BTS state.
  *
- * tracer: the tracer handle returned from ds_request_~()
- * index: the index of the requested record
- * record (out): pointer to the requested record
+ * This gives access to the raw DS state and adds functions to provide
+ * an arch-independent view of the BTS data.
  */
-extern int ds_access_bts(struct bts_tracer *tracer,
-			 size_t index, const void **record);
-extern int ds_access_pebs(struct pebs_tracer *tracer,
-			  size_t index, const void **record);
+struct bts_trace {
+	struct ds_trace ds;
+
+	int (*read)(struct bts_tracer *tracer, const void *at,
+		    struct bts_struct *out);
+	int (*write)(struct bts_tracer *tracer, const struct bts_struct *in);
+};
+
 
 /*
- * Write one or more BTS/PEBS records at the write pointer index and
- * advance the write pointer.
+ * The PEBS state.
  *
- * If size is not a multiple of the record size, trailing bytes are
- * zeroed out.
- *
- * May result in one or more overflow notifications.
- *
- * If called during overflow handling, that is, with index >=
- * interrupt threshold, the write will wrap around.
+ * This gives access to the raw DS state and the PEBS-specific counter
+ * reset value.
+ */
+struct pebs_trace {
+	struct ds_trace ds;
+
+	/* the PEBS reset value */
+	unsigned long long reset_value;
+};
+
+
+/*
+ * Read the BTS or PEBS trace.
  *
- * An overflow notification is given if and when the interrupt
- * threshold is reached during or after the write.
+ * Returns a view on the trace collected for the parameter tracer.
  *
- * Returns the number of bytes written or -Eerrno.
+ * The view remains valid as long as the traced task is not running or
+ * the tracer is suspended.
+ * Writes into the trace buffer are not reflected.
  *
  * tracer: the tracer handle returned from ds_request_~()
- * buffer: the buffer to write
- * size: the size of the buffer
  */
-extern int ds_write_bts(struct bts_tracer *tracer,
-			const void *buffer, size_t size);
-extern int ds_write_pebs(struct pebs_tracer *tracer,
-			 const void *buffer, size_t size);
+extern const struct bts_trace *ds_read_bts(struct bts_tracer *tracer);
+extern const struct pebs_trace *ds_read_pebs(struct pebs_tracer *tracer);
+
 
 /*
  * Reset the write pointer of the BTS/PEBS buffer.
@@ -155,27 +231,6 @@ extern int ds_write_pebs(struct pebs_tracer *tracer,
 extern int ds_reset_bts(struct bts_tracer *tracer);
 extern int ds_reset_pebs(struct pebs_tracer *tracer);
 
-/*
- * Clear the BTS/PEBS buffer and reset the write pointer.
- * The entire buffer will be zeroed out.
- *
- * Returns 0 on success; -Eerrno on error
- *
- * tracer: the tracer handle returned from ds_request_~()
- */
-extern int ds_clear_bts(struct bts_tracer *tracer);
-extern int ds_clear_pebs(struct pebs_tracer *tracer);
-
-/*
- * Provide the PEBS counter reset value.
- *
- * Returns 0 on success; -Eerrno on error
- *
- * tracer: the tracer handle returned from ds_request_pebs()
- * value (out): the counter reset value
- */
-extern int ds_get_pebs_reset(struct pebs_tracer *tracer, u64 *value);
-
 /*
  * Set the PEBS counter reset value.
  *
@@ -192,35 +247,17 @@ extern int ds_set_pebs_reset(struct pebs_tracer *tracer, u64 value);
 struct cpuinfo_x86;
 extern void __cpuinit ds_init_intel(struct cpuinfo_x86 *);
 
-
-
 /*
- * The DS context - part of struct thread_struct.
+ * Context switch work
  */
-#define MAX_SIZEOF_DS (12 * 8)
-
-struct ds_context {
-	/* pointer to the DS configuration; goes into MSR_IA32_DS_AREA */
-	unsigned char ds[MAX_SIZEOF_DS];
-	/* the owner of the BTS and PEBS configuration, respectively */
-	struct ds_tracer  *owner[2];
-	/* use count */
-	unsigned long count;
-	/* a pointer to the context location inside the thread_struct
-	 * or the per_cpu context array */
-	struct ds_context **this;
-	/* a pointer to the task owning this context, or NULL, if the
-	 * context is owned by a cpu */
-	struct task_struct *task;
-};
-
-/* called by exit_thread() to free leftover contexts */
-extern void ds_free(struct ds_context *context);
+extern void ds_switch_to(struct task_struct *prev, struct task_struct *next);
 
 #else /* CONFIG_X86_DS */
 
 struct cpuinfo_x86;
 static inline void __cpuinit ds_init_intel(struct cpuinfo_x86 *ignored) {}
+static inline void ds_switch_to(struct task_struct *prev,
+				struct task_struct *next) {}
 
 #endif /* CONFIG_X86_DS */
 #endif /* _ASM_X86_DS_H */

arch/x86/include/asm/processor.h

@@ -752,6 +752,19 @@ extern void switch_to_new_gdt(void);
 extern void cpu_init(void);
 extern void init_gdt(int cpu);
 
+static inline unsigned long get_debugctlmsr(void)
+{
+    unsigned long debugctlmsr = 0;
+
+#ifndef CONFIG_X86_DEBUGCTLMSR
+	if (boot_cpu_data.x86 < 6)
+		return 0;
+#endif
+	rdmsrl(MSR_IA32_DEBUGCTLMSR, debugctlmsr);
+
+    return debugctlmsr;
+}
+
 static inline void update_debugctlmsr(unsigned long debugctlmsr)
 {
 #ifndef CONFIG_X86_DEBUGCTLMSR

arch/x86/include/asm/ptrace.h

@@ -6,7 +6,6 @@
 #include <asm/processor-flags.h>
 
 #ifdef __KERNEL__
-#include <asm/ds.h>		/* the DS BTS struct is used for ptrace too */
 #include <asm/segment.h>
 #endif
 
@@ -128,34 +127,6 @@ struct pt_regs {
 #endif /* !__i386__ */
 
 
-#ifdef CONFIG_X86_PTRACE_BTS
-/* a branch trace record entry
- *
- * In order to unify the interface between various processor versions,
- * we use the below data structure for all processors.
- */
-enum bts_qualifier {
-	BTS_INVALID = 0,
-	BTS_BRANCH,
-	BTS_TASK_ARRIVES,
-	BTS_TASK_DEPARTS
-};
-
-struct bts_struct {
-	__u64 qualifier;
-	union {
-		/* BTS_BRANCH */
-		struct {
-			__u64 from_ip;
-			__u64 to_ip;
-		} lbr;
-		/* BTS_TASK_ARRIVES or
-		   BTS_TASK_DEPARTS */
-		__u64 jiffies;
-	} variant;
-};
-#endif /* CONFIG_X86_PTRACE_BTS */
-
 #ifdef __KERNEL__
 
 #include <linux/init.h>
@@ -163,13 +134,6 @@ struct bts_struct {
 struct cpuinfo_x86;
 struct task_struct;
 
-#ifdef CONFIG_X86_PTRACE_BTS
-extern void __cpuinit ptrace_bts_init_intel(struct cpuinfo_x86 *);
-extern void ptrace_bts_take_timestamp(struct task_struct *, enum bts_qualifier);
-#else
-#define ptrace_bts_init_intel(config) do {} while (0)
-#endif /* CONFIG_X86_PTRACE_BTS */
-
 extern unsigned long profile_pc(struct pt_regs *regs);
 
 extern unsigned long

arch/x86/include/asm/thread_info.h

@@ -93,7 +93,6 @@ struct thread_info {
 #define TIF_FORCED_TF		24	/* true if TF in eflags artificially */
 #define TIF_DEBUGCTLMSR		25	/* uses thread_struct.debugctlmsr */
 #define TIF_DS_AREA_MSR		26      /* uses thread_struct.ds_area_msr */
-#define TIF_BTS_TRACE_TS	27      /* record scheduling event timestamps */
 
 #define _TIF_SYSCALL_TRACE	(1 << TIF_SYSCALL_TRACE)
 #define _TIF_NOTIFY_RESUME	(1 << TIF_NOTIFY_RESUME)
@@ -115,7 +114,6 @@ struct thread_info {
 #define _TIF_FORCED_TF		(1 << TIF_FORCED_TF)
 #define _TIF_DEBUGCTLMSR	(1 << TIF_DEBUGCTLMSR)
 #define _TIF_DS_AREA_MSR	(1 << TIF_DS_AREA_MSR)
-#define _TIF_BTS_TRACE_TS	(1 << TIF_BTS_TRACE_TS)
 
 /* work to do in syscall_trace_enter() */
 #define _TIF_WORK_SYSCALL_ENTRY	\
@@ -141,8 +139,7 @@ struct thread_info {
 
 /* flags to check in __switch_to() */
 #define _TIF_WORK_CTXSW							\
-	(_TIF_IO_BITMAP|_TIF_DEBUGCTLMSR|_TIF_DS_AREA_MSR|_TIF_BTS_TRACE_TS| \
-								_TIF_NOTSC)
+	(_TIF_IO_BITMAP|_TIF_DEBUGCTLMSR|_TIF_DS_AREA_MSR|_TIF_NOTSC)
 
 #define _TIF_WORK_CTXSW_PREV _TIF_WORK_CTXSW
 #define _TIF_WORK_CTXSW_NEXT (_TIF_WORK_CTXSW|_TIF_DEBUG)

arch/x86/kernel/apic.c

@@ -30,6 +30,7 @@
 #include <linux/module.h>
 #include <linux/dmi.h>
 #include <linux/dmar.h>
+#include <linux/ftrace.h>
 
 #include <asm/atomic.h>
 #include <asm/smp.h>
@@ -800,7 +801,7 @@ static void local_apic_timer_interrupt(void)
  * [ if a single-CPU system runs an SMP kernel then we call the local
  *   interrupt as well. Thus we cannot inline the local irq ... ]
  */
-void smp_apic_timer_interrupt(struct pt_regs *regs)
+void __irq_entry smp_apic_timer_interrupt(struct pt_regs *regs)
 {
 	struct pt_regs *old_regs = set_irq_regs(regs);
 

arch/x86/kernel/cpu/intel.c

@@ -11,7 +11,6 @@
 #include <asm/pgtable.h>
 #include <asm/msr.h>
 #include <asm/uaccess.h>
-#include <asm/ptrace.h>
 #include <asm/ds.h>
 #include <asm/bugs.h>
 
@@ -309,9 +308,6 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
 		set_cpu_cap(c, X86_FEATURE_P3);
 #endif
 
-	if (cpu_has_bts)
-		ptrace_bts_init_intel(c);
-
 	detect_extended_topology(c);
 	if (!cpu_has(c, X86_FEATURE_XTOPOLOGY)) {
 		/*

arch/x86/kernel/irq_64.c

@@ -13,6 +13,7 @@
 #include <linux/seq_file.h>
 #include <linux/module.h>
 #include <linux/delay.h>
+#include <linux/ftrace.h>
 #include <asm/uaccess.h>
 #include <asm/io_apic.h>
 #include <asm/idle.h>
@@ -47,7 +48,7 @@ static inline void stack_overflow_check(struct pt_regs *regs)
  * SMP cross-CPU interrupts have their own specific
  * handlers).
  */
-asmlinkage unsigned int do_IRQ(struct pt_regs *regs)
+asmlinkage unsigned int __irq_entry do_IRQ(struct pt_regs *regs)
 {
 	struct pt_regs *old_regs = set_irq_regs(regs);
 	struct irq_desc *desc;

arch/x86/kernel/process_32.c

@@ -252,11 +252,14 @@ void exit_thread(void)
 		put_cpu();
 	}
 #ifdef CONFIG_X86_DS
-	/* Free any DS contexts that have not been properly released. */
-	if (unlikely(current->thread.ds_ctx)) {
-		/* we clear debugctl to make sure DS is not used. */
-		update_debugctlmsr(0);
-		ds_free(current->thread.ds_ctx);
+	/* Free any BTS tracers that have not been properly released. */
+	if (unlikely(current->bts)) {
+		ds_release_bts(current->bts);
+		current->bts = NULL;
+
+		kfree(current->bts_buffer);
+		current->bts_buffer = NULL;
+		current->bts_size = 0;
 	}
 #endif /* CONFIG_X86_DS */
 }
@@ -420,48 +423,19 @@ int set_tsc_mode(unsigned int val)
 	return 0;
 }
 
-#ifdef CONFIG_X86_DS
-static int update_debugctl(struct thread_struct *prev,
-			struct thread_struct *next, unsigned long debugctl)
-{
-	unsigned long ds_prev = 0;
-	unsigned long ds_next = 0;
-
-	if (prev->ds_ctx)
-		ds_prev = (unsigned long)prev->ds_ctx->ds;
-	if (next->ds_ctx)
-		ds_next = (unsigned long)next->ds_ctx->ds;
-
-	if (ds_next != ds_prev) {
-		/* we clear debugctl to make sure DS
-		 * is not in use when we change it */
-		debugctl = 0;
-		update_debugctlmsr(0);
-		wrmsr(MSR_IA32_DS_AREA, ds_next, 0);
-	}
-	return debugctl;
-}
-#else
-static int update_debugctl(struct thread_struct *prev,
-			struct thread_struct *next, unsigned long debugctl)
-{
-	return debugctl;
-}
-#endif /* CONFIG_X86_DS */
-
 static noinline void
 __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
 		 struct tss_struct *tss)
 {
 	struct thread_struct *prev, *next;
-	unsigned long debugctl;
 
 	prev = &prev_p->thread;
 	next = &next_p->thread;
 
-	debugctl = update_debugctl(prev, next, prev->debugctlmsr);
-
-	if (next->debugctlmsr != debugctl)
+	if (test_tsk_thread_flag(next_p, TIF_DS_AREA_MSR) ||
+	    test_tsk_thread_flag(prev_p, TIF_DS_AREA_MSR))
+		ds_switch_to(prev_p, next_p);
+	else if (next->debugctlmsr != prev->debugctlmsr)
 		update_debugctlmsr(next->debugctlmsr);
 
 	if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
@@ -483,15 +457,6 @@ __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
 			hard_enable_TSC();
 	}
 
-#ifdef CONFIG_X86_PTRACE_BTS
-	if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS))
-		ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS);
-
-	if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS))
-		ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES);
-#endif /* CONFIG_X86_PTRACE_BTS */
-
-
 	if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) {
 		/*
 		 * Disable the bitmap via an invalid offset. We still cache

arch/x86/kernel/process_64.c

@@ -237,11 +237,14 @@ void exit_thread(void)
 		put_cpu();
 	}
 #ifdef CONFIG_X86_DS
-	/* Free any DS contexts that have not been properly released. */
-	if (unlikely(t->ds_ctx)) {
-		/* we clear debugctl to make sure DS is not used. */
-		update_debugctlmsr(0);
-		ds_free(t->ds_ctx);
+	/* Free any BTS tracers that have not been properly released. */
+	if (unlikely(current->bts)) {
+		ds_release_bts(current->bts);
+		current->bts = NULL;
+
+		kfree(current->bts_buffer);
+		current->bts_buffer = NULL;
+		current->bts_size = 0;
 	}
 #endif /* CONFIG_X86_DS */
 }
@@ -471,35 +474,14 @@ static inline void __switch_to_xtra(struct task_struct *prev_p,
 				    struct tss_struct *tss)
 {
 	struct thread_struct *prev, *next;
-	unsigned long debugctl;
 
 	prev = &prev_p->thread,
 	next = &next_p->thread;
 
-	debugctl = prev->debugctlmsr;
-
-#ifdef CONFIG_X86_DS
-	{
-		unsigned long ds_prev = 0, ds_next = 0;
-
-		if (prev->ds_ctx)
-			ds_prev = (unsigned long)prev->ds_ctx->ds;
-		if (next->ds_ctx)
-			ds_next = (unsigned long)next->ds_ctx->ds;
-
-		if (ds_next != ds_prev) {
-			/*
-			 * We clear debugctl to make sure DS
-			 * is not in use when we change it:
-			 */
-			debugctl = 0;
-			update_debugctlmsr(0);
-			wrmsrl(MSR_IA32_DS_AREA, ds_next);
-		}
-	}
-#endif /* CONFIG_X86_DS */
-
-	if (next->debugctlmsr != debugctl)
+	if (test_tsk_thread_flag(next_p, TIF_DS_AREA_MSR) ||
+	    test_tsk_thread_flag(prev_p, TIF_DS_AREA_MSR))
+		ds_switch_to(prev_p, next_p);
+	else if (next->debugctlmsr != prev->debugctlmsr)
 		update_debugctlmsr(next->debugctlmsr);
 
 	if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
@@ -534,14 +516,6 @@ static inline void __switch_to_xtra(struct task_struct *prev_p,
 		 */
 		memset(tss->io_bitmap, 0xff, prev->io_bitmap_max);
 	}
-
-#ifdef CONFIG_X86_PTRACE_BTS
-	if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS))
-		ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS);
-
-	if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS))
-		ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES);
-#endif /* CONFIG_X86_PTRACE_BTS */
 }
 
 /*

arch/x86/kernel/vmlinux_32.lds.S

@@ -44,6 +44,7 @@ SECTIONS
 	SCHED_TEXT
 	LOCK_TEXT
 	KPROBES_TEXT
+	IRQENTRY_TEXT
 	*(.fixup)
 	*(.gnu.warning)
   	_etext = .;			/* End of text section */

arch/x86/kernel/vmlinux_64.lds.S

@@ -35,6 +35,7 @@ SECTIONS
 	SCHED_TEXT
 	LOCK_TEXT
 	KPROBES_TEXT
+	IRQENTRY_TEXT
 	*(.fixup)
 	*(.gnu.warning)
 	_etext = .;		/* End of text section */

include/asm-generic/vmlinux.lds.h

@@ -288,6 +288,16 @@
 		*(.kprobes.text)					\
 		VMLINUX_SYMBOL(__kprobes_text_end) = .;
 
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+#define IRQENTRY_TEXT							\
+		ALIGN_FUNCTION();					\
+		VMLINUX_SYMBOL(__irqentry_text_start) = .;		\
+		*(.irqentry.text)					\
+		VMLINUX_SYMBOL(__irqentry_text_end) = .;
+#else
+#define IRQENTRY_TEXT
+#endif
+
 /* Section used for early init (in .S files) */
 #define HEAD_TEXT  *(.head.text)
 

include/linux/ftrace.h

@@ -377,6 +377,16 @@ struct ftrace_graph_ret {
  */
 #define __notrace_funcgraph		notrace
 
+/*
+ * We want to which function is an entrypoint of a hardirq.
+ * That will help us to put a signal on output.
+ */
+#define __irq_entry		 __attribute__((__section__(".irqentry.text")))
+
+/* Limits of hardirq entrypoints */
+extern char __irqentry_text_start[];
+extern char __irqentry_text_end[];
+
 #define FTRACE_RETFUNC_DEPTH 50
 #define FTRACE_RETSTACK_ALLOC_SIZE 32
 /* Type of the callback handlers for tracing function graph*/
@@ -414,6 +424,7 @@ static inline void unpause_graph_tracing(void)
 #else
 
 #define __notrace_funcgraph
+#define __irq_entry
 
 static inline void ftrace_graph_init_task(struct task_struct *t) { }
 static inline void ftrace_graph_exit_task(struct task_struct *t) { }

include/linux/sched.h

@@ -1176,6 +1176,7 @@ struct task_struct {
 	 * The buffer to hold the BTS data.
 	 */
 	void *bts_buffer;
+	size_t bts_size;
 #endif /* CONFIG_X86_PTRACE_BTS */
 
 	/* PID/PID hash table linkage. */

include/trace/sched.h

@@ -30,8 +30,8 @@ DECLARE_TRACE(sched_switch,
 		TPARGS(rq, prev, next));
 
 DECLARE_TRACE(sched_migrate_task,
-	TPPROTO(struct rq *rq, struct task_struct *p, int dest_cpu),
-		TPARGS(rq, p, dest_cpu));
+	TPPROTO(struct task_struct *p, int orig_cpu, int dest_cpu),
+		TPARGS(p, orig_cpu, dest_cpu));
 
 DECLARE_TRACE(sched_process_free,
 	TPPROTO(struct task_struct *p),

kernel/sched.c

@@ -1851,6 +1851,8 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 
 	clock_offset = old_rq->clock - new_rq->clock;
 
+	trace_sched_migrate_task(p, task_cpu(p), new_cpu);
+
 #ifdef CONFIG_SCHEDSTATS
 	if (p->se.wait_start)
 		p->se.wait_start -= clock_offset;
@@ -2868,7 +2870,6 @@ static void sched_migrate_task(struct task_struct *p, int dest_cpu)
 	    || unlikely(!cpu_active(dest_cpu)))
 		goto out;
 
-	trace_sched_migrate_task(rq, p, dest_cpu);
 	/* force the process onto the specified CPU */
 	if (migrate_task(p, dest_cpu, &req)) {
 		/* Need to wait for migration thread (might exit: take ref). */

kernel/trace/Kconfig

@@ -251,9 +251,9 @@ config STACK_TRACER
 
 	  Say N if unsure.
 
-config BTS_TRACER
+config HW_BRANCH_TRACER
 	depends on HAVE_HW_BRANCH_TRACER
-	bool "Trace branches"
+	bool "Trace hw branches"
 	select TRACING
 	help
 	  This tracer records all branches on the system in a circular

kernel/trace/Makefile

@@ -31,7 +31,7 @@ obj-$(CONFIG_MMIOTRACE) += trace_mmiotrace.o
 obj-$(CONFIG_BOOT_TRACER) += trace_boot.o
 obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += trace_functions_graph.o
 obj-$(CONFIG_TRACE_BRANCH_PROFILING) += trace_branch.o
-obj-$(CONFIG_BTS_TRACER) += trace_bts.o
+obj-$(CONFIG_HW_BRANCH_TRACER) += trace_hw_branches.o
 obj-$(CONFIG_POWER_TRACER) += trace_power.o
 
 libftrace-y := ftrace.o

kernel/trace/trace.c

@@ -2425,7 +2425,7 @@ __tracing_open(struct inode *inode, struct file *file, int *ret)
 
 	/* Notify the tracer early; before we stop tracing. */
 	if (iter->trace && iter->trace->open)
-			iter->trace->open(iter);
+		iter->trace->open(iter);
 
 	/* Annotate start of buffers if we had overruns */
 	if (ring_buffer_overruns(iter->tr->buffer))

kernel/trace/trace.h

@@ -28,7 +28,7 @@ enum trace_type {
 	TRACE_GRAPH_RET,
 	TRACE_GRAPH_ENT,
 	TRACE_USER_STACK,
-	TRACE_BTS,
+	TRACE_HW_BRANCHES,
 	TRACE_POWER,
 
 	__TRACE_LAST_TYPE
@@ -159,10 +159,10 @@ struct trace_branch {
 	char			correct;
 };
 
-struct bts_entry {
+struct hw_branch_entry {
 	struct trace_entry	ent;
-	unsigned long		from;
-	unsigned long		to;
+	u64			from;
+	u64			to;
 };
 
 struct trace_power {
@@ -278,7 +278,7 @@ extern void __ftrace_bad_type(void);
 			  TRACE_GRAPH_ENT);		\
 		IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,	\
 			  TRACE_GRAPH_RET);		\
-		IF_ASSIGN(var, ent, struct bts_entry, TRACE_BTS);\
+		IF_ASSIGN(var, ent, struct hw_branch_entry, TRACE_HW_BRANCHES);\
  		IF_ASSIGN(var, ent, struct trace_power, TRACE_POWER); \
 		__ftrace_bad_type();					\
 	} while (0)
@@ -414,9 +414,7 @@ void trace_function(struct trace_array *tr,
 
 void trace_graph_return(struct ftrace_graph_ret *trace);
 int trace_graph_entry(struct ftrace_graph_ent *trace);
-void trace_bts(struct trace_array *tr,
-	       unsigned long from,
-	       unsigned long to);
+void trace_hw_branch(struct trace_array *tr, u64 from, u64 to);
 
 void tracing_start_cmdline_record(void);
 void tracing_stop_cmdline_record(void);

kernel/trace/trace_functions_graph.c

@@ -231,6 +231,49 @@ trace_branch_is_leaf(struct trace_iterator *iter,
 	return true;
 }
 
+static enum print_line_t
+print_graph_irq(struct trace_seq *s, unsigned long addr,
+				enum trace_type type, int cpu, pid_t pid)
+{
+	int ret;
+
+	if (addr < (unsigned long)__irqentry_text_start ||
+		addr >= (unsigned long)__irqentry_text_end)
+		return TRACE_TYPE_UNHANDLED;
+
+	if (type == TRACE_GRAPH_ENT) {
+		ret = trace_seq_printf(s, "==========> |  ");
+	} else {
+		/* Cpu */
+		if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU) {
+			ret = print_graph_cpu(s, cpu);
+			if (ret == TRACE_TYPE_PARTIAL_LINE)
+				return TRACE_TYPE_PARTIAL_LINE;
+		}
+		/* Proc */
+		if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC) {
+			ret = print_graph_proc(s, pid);
+			if (ret == TRACE_TYPE_PARTIAL_LINE)
+				return TRACE_TYPE_PARTIAL_LINE;
+
+			ret = trace_seq_printf(s, " | ");
+			if (!ret)
+				return TRACE_TYPE_PARTIAL_LINE;
+		}
+
+		/* No overhead */
+		if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERHEAD) {
+			ret = trace_seq_printf(s, "  ");
+			if (!ret)
+				return TRACE_TYPE_PARTIAL_LINE;
+		}
+
+		ret = trace_seq_printf(s, "<========== |\n");
+	}
+	if (!ret)
+		return TRACE_TYPE_PARTIAL_LINE;
+	return TRACE_TYPE_HANDLED;
+}
 
 static enum print_line_t
 print_graph_duration(unsigned long long duration, struct trace_seq *s)
@@ -344,7 +387,7 @@ print_graph_entry_leaf(struct trace_iterator *iter,
 
 static enum print_line_t
 print_graph_entry_nested(struct ftrace_graph_ent_entry *entry,
-			struct trace_seq *s)
+			struct trace_seq *s, pid_t pid, int cpu)
 {
 	int i;
 	int ret;
@@ -357,8 +400,18 @@ print_graph_entry_nested(struct ftrace_graph_ent_entry *entry,
 			return TRACE_TYPE_PARTIAL_LINE;
 	}
 
-	/* No time */
-	ret = trace_seq_printf(s, "            |  ");
+	/* Interrupt */
+	ret = print_graph_irq(s, call->func, TRACE_GRAPH_ENT, cpu, pid);
+	if (ret == TRACE_TYPE_UNHANDLED) {
+		/* No time */
+		ret = trace_seq_printf(s, "            |  ");
+		if (!ret)
+			return TRACE_TYPE_PARTIAL_LINE;
+	} else {
+		if (ret == TRACE_TYPE_PARTIAL_LINE)
+			return TRACE_TYPE_PARTIAL_LINE;
+	}
+
 
 	/* Function */
 	for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
@@ -410,7 +463,7 @@ print_graph_entry(struct ftrace_graph_ent_entry *field, struct trace_seq *s,
 	if (trace_branch_is_leaf(iter, field))
 		return print_graph_entry_leaf(iter, field, s);
 	else
-		return print_graph_entry_nested(field, s);
+		return print_graph_entry_nested(field, s, iter->ent->pid, cpu);
 
 }
 
@@ -474,6 +527,11 @@ print_graph_return(struct ftrace_graph_ret *trace, struct trace_seq *s,
 		if (!ret)
 			return TRACE_TYPE_PARTIAL_LINE;
 	}
+
+	ret = print_graph_irq(s, trace->func, TRACE_GRAPH_RET, cpu, ent->pid);
+	if (ret == TRACE_TYPE_PARTIAL_LINE)
+		return TRACE_TYPE_PARTIAL_LINE;
+
 	return TRACE_TYPE_HANDLED;
 }
 

kernel/trace/trace_bts.c → kernel/trace/trace_hw_branches.c

 /*
- * BTS tracer
+ * h/w branch tracer for x86 based on bts
  *
  * Copyright (C) 2008 Markus Metzger <markus.t.metzger@gmail.com>
  *
@@ -25,68 +25,6 @@ static DEFINE_PER_CPU(unsigned char[SIZEOF_BTS], buffer);
 #define this_buffer per_cpu(buffer, smp_processor_id())
 
 
-/*
- * Information to interpret a BTS record.
- * This will go into an in-kernel BTS interface.
- */
-static unsigned char sizeof_field;
-static unsigned long debugctl_mask;
-
-#define sizeof_bts (3 * sizeof_field)
-
-static void bts_trace_cpuinit(struct cpuinfo_x86 *c)
-{
-	switch (c->x86) {
-	case 0x6:
-		switch (c->x86_model) {
-		case 0x0 ... 0xC:
-			break;
-		case 0xD:
-		case 0xE: /* Pentium M */
-			sizeof_field = sizeof(long);
-			debugctl_mask = (1<<6)|(1<<7);
-			break;
-		default:
-			sizeof_field = 8;
-			debugctl_mask = (1<<6)|(1<<7);
-			break;
-		}
-		break;
-	case 0xF:
-		switch (c->x86_model) {
-		case 0x0:
-		case 0x1:
-		case 0x2: /* Netburst */
-			sizeof_field = sizeof(long);
-			debugctl_mask = (1<<2)|(1<<3);
-			break;
-		default:
-			/* sorry, don't know about them */
-			break;
-		}
-		break;
-	default:
-		/* sorry, don't know about them */
-		break;
-	}
-}
-
-static inline void bts_enable(void)
-{
-	unsigned long debugctl;
-
-	rdmsrl(MSR_IA32_DEBUGCTLMSR, debugctl);
-	wrmsrl(MSR_IA32_DEBUGCTLMSR, debugctl | debugctl_mask);
-}
-
-static inline void bts_disable(void)
-{
-	unsigned long debugctl;
-
-	rdmsrl(MSR_IA32_DEBUGCTLMSR, debugctl);
-	wrmsrl(MSR_IA32_DEBUGCTLMSR, debugctl & ~debugctl_mask);
-}
-
 static void bts_trace_reset(struct trace_array *tr)
 {
 	int cpu;
@@ -99,15 +37,17 @@ static void bts_trace_reset(struct trace_array *tr)
 
 static void bts_trace_start_cpu(void *arg)
 {
+	if (this_tracer)
+		ds_release_bts(this_tracer);
+
 	this_tracer =
 		ds_request_bts(/* task = */ NULL, this_buffer, SIZEOF_BTS,
-			       /* ovfl = */ NULL, /* th = */ (size_t)-1);
+			       /* ovfl = */ NULL, /* th = */ (size_t)-1,
+			       BTS_KERNEL);
 	if (IS_ERR(this_tracer)) {
 		this_tracer = NULL;
 		return;
 	}
-
-	bts_enable();
 }
 
 static void bts_trace_start(struct trace_array *tr)
@@ -123,8 +63,6 @@ static void bts_trace_start(struct trace_array *tr)
 static void bts_trace_stop_cpu(void *arg)
 {
 	if (this_tracer) {
-		bts_disable();
-
 		ds_release_bts(this_tracer);
 		this_tracer = NULL;
 	}
@@ -140,7 +78,6 @@ static void bts_trace_stop(struct trace_array *tr)
 
 static int bts_trace_init(struct trace_array *tr)
 {
-	bts_trace_cpuinit(&boot_cpu_data);
 	bts_trace_reset(tr);
 	bts_trace_start(tr);
 
@@ -149,47 +86,37 @@ static int bts_trace_init(struct trace_array *tr)
 
 static void bts_trace_print_header(struct seq_file *m)
 {
-#ifdef __i386__
-	seq_puts(m, "# CPU#    FROM           TO     FUNCTION\n");
-	seq_puts(m, "#  |       |             |         |\n");
-#else
 	seq_puts(m,
 		 "# CPU#        FROM                   TO         FUNCTION\n");
 	seq_puts(m,
 		 "#  |           |                     |             |\n");
-#endif
 }
 
 static enum print_line_t bts_trace_print_line(struct trace_iterator *iter)
 {
 	struct trace_entry *entry = iter->ent;
 	struct trace_seq *seq = &iter->seq;
-	struct bts_entry *it;
+	struct hw_branch_entry *it;
 
 	trace_assign_type(it, entry);
 
-	if (entry->type == TRACE_BTS) {
-		int ret;
-#ifdef CONFIG_KALLSYMS
-		char function[KSYM_SYMBOL_LEN];
-		sprint_symbol(function, it->from);
-#else
-		char *function = "<unknown>";
-#endif
-
-		ret = trace_seq_printf(seq, "%4d  0x%lx -> 0x%lx [%s]\n",
-				       entry->cpu, it->from, it->to, function);
-		if (!ret)
-			return TRACE_TYPE_PARTIAL_LINE;;
-		return TRACE_TYPE_HANDLED;
+	if (entry->type == TRACE_HW_BRANCHES) {
+		if (trace_seq_printf(seq, "%4d  ", entry->cpu) &&
+		    trace_seq_printf(seq, "0x%016llx -> 0x%016llx ",
+				     it->from, it->to) &&
+		    (!it->from ||
+		     seq_print_ip_sym(seq, it->from, /* sym_flags = */ 0)) &&
+		    trace_seq_printf(seq, "\n"))
+			return TRACE_TYPE_HANDLED;
+		return TRACE_TYPE_PARTIAL_LINE;;
 	}
 	return TRACE_TYPE_UNHANDLED;
 }
 
-void trace_bts(struct trace_array *tr, unsigned long from, unsigned long to)
+void trace_hw_branch(struct trace_array *tr, u64 from, u64 to)
 {
 	struct ring_buffer_event *event;
-	struct bts_entry *entry;
+	struct hw_branch_entry *entry;
 	unsigned long irq;
 
 	event = ring_buffer_lock_reserve(tr->buffer, sizeof(*entry), &irq);
@@ -197,56 +124,58 @@ void trace_bts(struct trace_array *tr, unsigned long from, unsigned long to)
 		return;
 	entry	= ring_buffer_event_data(event);
 	tracing_generic_entry_update(&entry->ent, 0, from);
-	entry->ent.type = TRACE_BTS;
+	entry->ent.type = TRACE_HW_BRANCHES;
 	entry->ent.cpu = smp_processor_id();
 	entry->from = from;
 	entry->to   = to;
 	ring_buffer_unlock_commit(tr->buffer, event, irq);
 }
 
-static void trace_bts_at(struct trace_array *tr, size_t index)
+static void trace_bts_at(struct trace_array *tr,
+			 const struct bts_trace *trace, void *at)
 {
-	const void *raw = NULL;
-	unsigned long from, to;
-	int err;
+	struct bts_struct bts;
+	int err = 0;
 
-	err = ds_access_bts(this_tracer, index, &raw);
-	if (err < 0)
+	WARN_ON_ONCE(!trace->read);
+	if (!trace->read)
 		return;
 
-	from = *(const unsigned long *)raw;
-	to = *(const unsigned long *)((const char *)raw + sizeof_field);
+	err = trace->read(this_tracer, at, &bts);
+	if (err < 0)
+		return;
 
-	trace_bts(tr, from, to);
+	switch (bts.qualifier) {
+	case BTS_BRANCH:
+		trace_hw_branch(tr, bts.variant.lbr.from, bts.variant.lbr.to);
+		break;
+	}
 }
 
 static void trace_bts_cpu(void *arg)
 {
 	struct trace_array *tr = (struct trace_array *) arg;
-	size_t index = 0, end = 0, i;
-	int err;
+	const struct bts_trace *trace;
+	unsigned char *at;
 
 	if (!this_tracer)
 		return;
 
-	bts_disable();
-
-	err = ds_get_bts_index(this_tracer, &index);
-	if (err < 0)
-		goto out;
-
-	err = ds_get_bts_end(this_tracer, &end);
-	if (err < 0)
+	ds_suspend_bts(this_tracer);
+	trace = ds_read_bts(this_tracer);
+	if (!trace)
 		goto out;
 
-	for (i = index; i < end; i++)
-		trace_bts_at(tr, i);
+	for (at = trace->ds.top; (void *)at < trace->ds.end;
+	     at += trace->ds.size)
+		trace_bts_at(tr, trace, at);
 
-	for (i = 0; i < index; i++)
-		trace_bts_at(tr, i);
+	for (at = trace->ds.begin; (void *)at < trace->ds.top;
+	     at += trace->ds.size)
+		trace_bts_at(tr, trace, at);
 
 out:
-	bts_enable();
+	ds_resume_bts(this_tracer);
 }
 
 static void trace_bts_prepare(struct trace_iterator *iter)
@@ -259,7 +188,7 @@ static void trace_bts_prepare(struct trace_iterator *iter)
 
 struct tracer bts_tracer __read_mostly =
 {
-	.name		= "bts",
+	.name		= "hw-branch-tracer",
 	.init		= bts_trace_init,
 	.reset		= bts_trace_stop,
 	.print_header	= bts_trace_print_header,

scripts/recordmcount.pl

@@ -114,6 +114,7 @@ my %text_sections = (
      ".text" => 1,
      ".sched.text" => 1,
      ".spinlock.text" => 1,
+     ".irqentry.text" => 1,
 );
 
 $objdump = "objdump" if ((length $objdump) == 0);