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: (50 commits)
  perf python scripting: Add futex-contention script
  perf python scripting: Fixup cut'n'paste error in sctop script
  perf scripting: Shut up 'perf record' final status
  perf record: Remove newline character from perror() argument
  perf python scripting: Support fedora 11 (audit 1.7.17)
  perf python scripting: Improve the syscalls-by-pid script
  perf python scripting: print the syscall name on sctop
  perf python scripting: Improve the syscalls-counts script
  perf python scripting: Improve the failed-syscalls-by-pid script
  kprobes: Remove redundant text_mutex lock in optimize
  x86/oprofile: Fix uninitialized variable use in debug printk
  tracing: Fix 'faild' -> 'failed' typo
  perf probe: Fix format specified for Dwarf_Off parameter
  perf trace: Fix detection of script extension
  perf trace: Use $PERF_EXEC_PATH in canned report scripts
  perf tools: Document event modifiers
  perf tools: Remove direct slang.h include
  perf_events: Fix for transaction recovery in group_sched_in()
  perf_events: Revert: Fix transaction recovery in group_sched_in()
  perf, x86: Use NUMA aware allocations for PEBS/BTS/DS allocations
  ...

arch/x86/include/asm/msr-index.h

@@ -121,6 +121,7 @@
 #define MSR_AMD64_IBSDCLINAD		0xc0011038
 #define MSR_AMD64_IBSDCPHYSAD		0xc0011039
 #define MSR_AMD64_IBSCTL		0xc001103a
+#define MSR_AMD64_IBSBRTARGET		0xc001103b
 
 /* Fam 10h MSRs */
 #define MSR_FAM10H_MMIO_CONF_BASE	0xc0010058

arch/x86/include/asm/perf_event.h

@@ -122,6 +122,7 @@ union cpuid10_edx {
 #define IBS_OP_VAL		(1ULL<<18)
 #define IBS_OP_ENABLE		(1ULL<<17)
 #define IBS_OP_MAX_CNT		0x0000FFFFULL
+#define IBS_OP_MAX_CNT_EXT	0x007FFFFFULL	/* not a register bit mask */
 
 #ifdef CONFIG_PERF_EVENTS
 extern void init_hw_perf_events(void);

arch/x86/kernel/cpu/perf_event.c

@@ -237,6 +237,7 @@ struct x86_pmu {
 	 * Intel DebugStore bits
 	 */
 	int		bts, pebs;
+	int		bts_active, pebs_active;
 	int		pebs_record_size;
 	void		(*drain_pebs)(struct pt_regs *regs);
 	struct event_constraint *pebs_constraints;
@@ -380,7 +381,7 @@ static void release_pmc_hardware(void) {}
 
 #endif
 
-static int reserve_ds_buffers(void);
+static void reserve_ds_buffers(void);
 static void release_ds_buffers(void);
 
 static void hw_perf_event_destroy(struct perf_event *event)
@@ -477,7 +478,7 @@ static int x86_setup_perfctr(struct perf_event *event)
 	if ((attr->config == PERF_COUNT_HW_BRANCH_INSTRUCTIONS) &&
 	    (hwc->sample_period == 1)) {
 		/* BTS is not supported by this architecture. */
-		if (!x86_pmu.bts)
+		if (!x86_pmu.bts_active)
 			return -EOPNOTSUPP;
 
 		/* BTS is currently only allowed for user-mode. */
@@ -496,12 +497,13 @@ static int x86_pmu_hw_config(struct perf_event *event)
 		int precise = 0;
 
 		/* Support for constant skid */
-		if (x86_pmu.pebs)
+		if (x86_pmu.pebs_active) {
 			precise++;
 
 			/* Support for IP fixup */
 			if (x86_pmu.lbr_nr)
 				precise++;
+		}
 
 		if (event->attr.precise_ip > precise)
 			return -EOPNOTSUPP;
@@ -543,11 +545,8 @@ static int __x86_pmu_event_init(struct perf_event *event)
 		if (atomic_read(&active_events) == 0) {
 			if (!reserve_pmc_hardware())
 				err = -EBUSY;
-			else {
-				err = reserve_ds_buffers();
-				if (err)
-					release_pmc_hardware();
-			}
+			else
+				reserve_ds_buffers();
 		}
 		if (!err)
 			atomic_inc(&active_events);

arch/x86/kernel/cpu/perf_event_intel_ds.c

@@ -74,58 +74,57 @@ static void fini_debug_store_on_cpu(int cpu)
 	wrmsr_on_cpu(cpu, MSR_IA32_DS_AREA, 0, 0);
 }
 
-static void release_ds_buffers(void)
+static int alloc_pebs_buffer(int cpu)
 {
-	int cpu;
-
-	if (!x86_pmu.bts && !x86_pmu.pebs)
-		return;
-
-	get_online_cpus();
+	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
+	int node = cpu_to_node(cpu);
+	int max, thresh = 1; /* always use a single PEBS record */
+	void *buffer;
 
-	for_each_online_cpu(cpu)
-		fini_debug_store_on_cpu(cpu);
+	if (!x86_pmu.pebs)
+		return 0;
 
-	for_each_possible_cpu(cpu) {
-		struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
+	buffer = kmalloc_node(PEBS_BUFFER_SIZE, GFP_KERNEL | __GFP_ZERO, node);
+	if (unlikely(!buffer))
+		return -ENOMEM;
 
-		if (!ds)
-			continue;
+	max = PEBS_BUFFER_SIZE / x86_pmu.pebs_record_size;
 
-		per_cpu(cpu_hw_events, cpu).ds = NULL;
+	ds->pebs_buffer_base = (u64)(unsigned long)buffer;
+	ds->pebs_index = ds->pebs_buffer_base;
+	ds->pebs_absolute_maximum = ds->pebs_buffer_base +
+		max * x86_pmu.pebs_record_size;
 
-		kfree((void *)(unsigned long)ds->pebs_buffer_base);
-		kfree((void *)(unsigned long)ds->bts_buffer_base);
-		kfree(ds);
-	}
+	ds->pebs_interrupt_threshold = ds->pebs_buffer_base +
+		thresh * x86_pmu.pebs_record_size;
 
-	put_online_cpus();
+	return 0;
 }
 
-static int reserve_ds_buffers(void)
+static void release_pebs_buffer(int cpu)
 {
-	int cpu, err = 0;
+	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
 
-	if (!x86_pmu.bts && !x86_pmu.pebs)
-		return 0;
+	if (!ds || !x86_pmu.pebs)
+		return;
 
-	get_online_cpus();
+	kfree((void *)(unsigned long)ds->pebs_buffer_base);
+	ds->pebs_buffer_base = 0;
+}
 
-	for_each_possible_cpu(cpu) {
-		struct debug_store *ds;
-		void *buffer;
+static int alloc_bts_buffer(int cpu)
+{
+	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
+	int node = cpu_to_node(cpu);
 	int max, thresh;
+	void *buffer;
 
-		err = -ENOMEM;
-		ds = kzalloc(sizeof(*ds), GFP_KERNEL);
-		if (unlikely(!ds))
-			break;
-		per_cpu(cpu_hw_events, cpu).ds = ds;
+	if (!x86_pmu.bts)
+		return 0;
 
-		if (x86_pmu.bts) {
-			buffer = kzalloc(BTS_BUFFER_SIZE, GFP_KERNEL);
+	buffer = kmalloc_node(BTS_BUFFER_SIZE, GFP_KERNEL | __GFP_ZERO, node);
 	if (unlikely(!buffer))
-				break;
+		return -ENOMEM;
 
 	max = BTS_BUFFER_SIZE / BTS_RECORD_SIZE;
 	thresh = max / 16;
@@ -136,39 +135,125 @@ static int reserve_ds_buffers(void)
 		max * BTS_RECORD_SIZE;
 	ds->bts_interrupt_threshold = ds->bts_absolute_maximum -
 		thresh * BTS_RECORD_SIZE;
+
+	return 0;
+}
+
+static void release_bts_buffer(int cpu)
+{
+	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
+
+	if (!ds || !x86_pmu.bts)
+		return;
+
+	kfree((void *)(unsigned long)ds->bts_buffer_base);
+	ds->bts_buffer_base = 0;
+}
+
+static int alloc_ds_buffer(int cpu)
+{
+	int node = cpu_to_node(cpu);
+	struct debug_store *ds;
+
+	ds = kmalloc_node(sizeof(*ds), GFP_KERNEL | __GFP_ZERO, node);
+	if (unlikely(!ds))
+		return -ENOMEM;
+
+	per_cpu(cpu_hw_events, cpu).ds = ds;
+
+	return 0;
+}
+
+static void release_ds_buffer(int cpu)
+{
+	struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
+
+	if (!ds)
+		return;
+
+	per_cpu(cpu_hw_events, cpu).ds = NULL;
+	kfree(ds);
+}
+
+static void release_ds_buffers(void)
+{
+	int cpu;
+
+	if (!x86_pmu.bts && !x86_pmu.pebs)
+		return;
+
+	get_online_cpus();
+	for_each_online_cpu(cpu)
+		fini_debug_store_on_cpu(cpu);
+
+	for_each_possible_cpu(cpu) {
+		release_pebs_buffer(cpu);
+		release_bts_buffer(cpu);
+		release_ds_buffer(cpu);
 	}
+	put_online_cpus();
+}
 
-		if (x86_pmu.pebs) {
-			buffer = kzalloc(PEBS_BUFFER_SIZE, GFP_KERNEL);
-			if (unlikely(!buffer))
-				break;
+static void reserve_ds_buffers(void)
+{
+	int bts_err = 0, pebs_err = 0;
+	int cpu;
 
-			max = PEBS_BUFFER_SIZE / x86_pmu.pebs_record_size;
+	x86_pmu.bts_active = 0;
+	x86_pmu.pebs_active = 0;
 
-			ds->pebs_buffer_base = (u64)(unsigned long)buffer;
-			ds->pebs_index = ds->pebs_buffer_base;
-			ds->pebs_absolute_maximum = ds->pebs_buffer_base +
-				max * x86_pmu.pebs_record_size;
-			/*
-			 * Always use single record PEBS
-			 */
-			ds->pebs_interrupt_threshold = ds->pebs_buffer_base +
-				x86_pmu.pebs_record_size;
+	if (!x86_pmu.bts && !x86_pmu.pebs)
+		return;
+
+	if (!x86_pmu.bts)
+		bts_err = 1;
+
+	if (!x86_pmu.pebs)
+		pebs_err = 1;
+
+	get_online_cpus();
+
+	for_each_possible_cpu(cpu) {
+		if (alloc_ds_buffer(cpu)) {
+			bts_err = 1;
+			pebs_err = 1;
+		}
+
+		if (!bts_err && alloc_bts_buffer(cpu))
+			bts_err = 1;
+
+		if (!pebs_err && alloc_pebs_buffer(cpu))
+			pebs_err = 1;
+
+		if (bts_err && pebs_err)
+			break;
+	}
+
+	if (bts_err) {
+		for_each_possible_cpu(cpu)
+			release_bts_buffer(cpu);
 	}
 
-		err = 0;
+	if (pebs_err) {
+		for_each_possible_cpu(cpu)
+			release_pebs_buffer(cpu);
 	}
 
-	if (err)
-		release_ds_buffers();
-	else {
+	if (bts_err && pebs_err) {
+		for_each_possible_cpu(cpu)
+			release_ds_buffer(cpu);
+	} else {
+		if (x86_pmu.bts && !bts_err)
+			x86_pmu.bts_active = 1;
+
+		if (x86_pmu.pebs && !pebs_err)
+			x86_pmu.pebs_active = 1;
+
 		for_each_online_cpu(cpu)
 			init_debug_store_on_cpu(cpu);
 	}
 
 	put_online_cpus();
-
-	return err;
 }
 
 /*
@@ -233,7 +318,7 @@ static int intel_pmu_drain_bts_buffer(void)
 	if (!event)
 		return 0;
 
-	if (!ds)
+	if (!x86_pmu.bts_active)
 		return 0;
 
 	at  = (struct bts_record *)(unsigned long)ds->bts_buffer_base;
@@ -503,7 +588,7 @@ static void intel_pmu_drain_pebs_core(struct pt_regs *iregs)
 	struct pebs_record_core *at, *top;
 	int n;
 
-	if (!ds || !x86_pmu.pebs)
+	if (!x86_pmu.pebs_active)
 		return;
 
 	at  = (struct pebs_record_core *)(unsigned long)ds->pebs_buffer_base;
@@ -545,7 +630,7 @@ static void intel_pmu_drain_pebs_nhm(struct pt_regs *iregs)
 	u64 status = 0;
 	int bit, n;
 
-	if (!ds || !x86_pmu.pebs)
+	if (!x86_pmu.pebs_active)
 		return;
 
 	at  = (struct pebs_record_nhm *)(unsigned long)ds->pebs_buffer_base;
@@ -630,9 +715,8 @@ static void intel_ds_init(void)
 
 #else /* CONFIG_CPU_SUP_INTEL */
 
-static int reserve_ds_buffers(void)
+static void reserve_ds_buffers(void)
 {
-	return 0;
 }
 
 static void release_ds_buffers(void)

arch/x86/oprofile/nmi_int.c

@@ -726,6 +726,12 @@ int __init op_nmi_init(struct oprofile_operations *ops)
 		case 0x11:
 			cpu_type = "x86-64/family11h";
 			break;
+		case 0x12:
+			cpu_type = "x86-64/family12h";
+			break;
+		case 0x14:
+			cpu_type = "x86-64/family14h";
+			break;
 		default:
 			return -ENODEV;
 		}

arch/x86/oprofile/op_model_amd.c

@@ -48,17 +48,24 @@ static unsigned long reset_value[NUM_VIRT_COUNTERS];
 
 static u32 ibs_caps;
 
-struct op_ibs_config {
+struct ibs_config {
 	unsigned long op_enabled;
 	unsigned long fetch_enabled;
 	unsigned long max_cnt_fetch;
 	unsigned long max_cnt_op;
 	unsigned long rand_en;
 	unsigned long dispatched_ops;
+	unsigned long branch_target;
 };
 
-static struct op_ibs_config ibs_config;
-static u64 ibs_op_ctl;
+struct ibs_state {
+	u64		ibs_op_ctl;
+	int		branch_target;
+	unsigned long	sample_size;
+};
+
+static struct ibs_config ibs_config;
+static struct ibs_state ibs_state;
 
 /*
  * IBS cpuid feature detection
@@ -71,8 +78,16 @@ static u64 ibs_op_ctl;
  * bit 0 is used to indicate the existence of IBS.
  */
 #define IBS_CAPS_AVAIL			(1U<<0)
+#define IBS_CAPS_FETCHSAM		(1U<<1)
+#define IBS_CAPS_OPSAM			(1U<<2)
 #define IBS_CAPS_RDWROPCNT		(1U<<3)
 #define IBS_CAPS_OPCNT			(1U<<4)
+#define IBS_CAPS_BRNTRGT		(1U<<5)
+#define IBS_CAPS_OPCNTEXT		(1U<<6)
+
+#define IBS_CAPS_DEFAULT		(IBS_CAPS_AVAIL		\
+					 | IBS_CAPS_FETCHSAM	\
+					 | IBS_CAPS_OPSAM)
 
 /*
  * IBS APIC setup
@@ -99,12 +114,12 @@ static u32 get_ibs_caps(void)
 	/* check IBS cpuid feature flags */
 	max_level = cpuid_eax(0x80000000);
 	if (max_level < IBS_CPUID_FEATURES)
-		return IBS_CAPS_AVAIL;
+		return IBS_CAPS_DEFAULT;
 
 	ibs_caps = cpuid_eax(IBS_CPUID_FEATURES);
 	if (!(ibs_caps & IBS_CAPS_AVAIL))
 		/* cpuid flags not valid */
-		return IBS_CAPS_AVAIL;
+		return IBS_CAPS_DEFAULT;
 
 	return ibs_caps;
 }
@@ -197,8 +212,8 @@ op_amd_handle_ibs(struct pt_regs * const regs,
 		rdmsrl(MSR_AMD64_IBSOPCTL, ctl);
 		if (ctl & IBS_OP_VAL) {
 			rdmsrl(MSR_AMD64_IBSOPRIP, val);
-			oprofile_write_reserve(&entry, regs, val,
-					       IBS_OP_CODE, IBS_OP_SIZE);
+			oprofile_write_reserve(&entry, regs, val, IBS_OP_CODE,
+					       ibs_state.sample_size);
 			oprofile_add_data64(&entry, val);
 			rdmsrl(MSR_AMD64_IBSOPDATA, val);
 			oprofile_add_data64(&entry, val);
@@ -210,10 +225,14 @@ op_amd_handle_ibs(struct pt_regs * const regs,
 			oprofile_add_data64(&entry, val);
 			rdmsrl(MSR_AMD64_IBSDCPHYSAD, val);
 			oprofile_add_data64(&entry, val);
+			if (ibs_state.branch_target) {
+				rdmsrl(MSR_AMD64_IBSBRTARGET, val);
+				oprofile_add_data(&entry, (unsigned long)val);
+			}
 			oprofile_write_commit(&entry);
 
 			/* reenable the IRQ */
-			ctl = op_amd_randomize_ibs_op(ibs_op_ctl);
+			ctl = op_amd_randomize_ibs_op(ibs_state.ibs_op_ctl);
 			wrmsrl(MSR_AMD64_IBSOPCTL, ctl);
 		}
 	}
@@ -226,21 +245,32 @@ static inline void op_amd_start_ibs(void)
 	if (!ibs_caps)
 		return;
 
+	memset(&ibs_state, 0, sizeof(ibs_state));
+
+	/*
+	 * Note: Since the max count settings may out of range we
+	 * write back the actual used values so that userland can read
+	 * it.
+	 */
+
 	if (ibs_config.fetch_enabled) {
-		val = (ibs_config.max_cnt_fetch >> 4) & IBS_FETCH_MAX_CNT;
+		val = ibs_config.max_cnt_fetch >> 4;
+		val = min(val, IBS_FETCH_MAX_CNT);
+		ibs_config.max_cnt_fetch = val << 4;
 		val |= ibs_config.rand_en ? IBS_FETCH_RAND_EN : 0;
 		val |= IBS_FETCH_ENABLE;
 		wrmsrl(MSR_AMD64_IBSFETCHCTL, val);
 	}
 
 	if (ibs_config.op_enabled) {
-		ibs_op_ctl = ibs_config.max_cnt_op >> 4;
+		val = ibs_config.max_cnt_op >> 4;
 		if (!(ibs_caps & IBS_CAPS_RDWROPCNT)) {
 			/*
 			 * IbsOpCurCnt not supported.  See
 			 * op_amd_randomize_ibs_op() for details.
 			 */
-			ibs_op_ctl = clamp(ibs_op_ctl, 0x0081ULL, 0xFF80ULL);
+			val = clamp(val, 0x0081ULL, 0xFF80ULL);
+			ibs_config.max_cnt_op = val << 4;
 		} else {
 			/*
 			 * The start value is randomized with a
@@ -248,13 +278,24 @@ static inline void op_amd_start_ibs(void)
 			 * with the half of the randomized range. Also
 			 * avoid underflows.
 			 */
-			ibs_op_ctl = min(ibs_op_ctl + IBS_RANDOM_MAXCNT_OFFSET,
-					 IBS_OP_MAX_CNT);
+			val += IBS_RANDOM_MAXCNT_OFFSET;
+			if (ibs_caps & IBS_CAPS_OPCNTEXT)
+				val = min(val, IBS_OP_MAX_CNT_EXT);
+			else
+				val = min(val, IBS_OP_MAX_CNT);
+			ibs_config.max_cnt_op =
+				(val - IBS_RANDOM_MAXCNT_OFFSET) << 4;
+		}
+		val = ((val & ~IBS_OP_MAX_CNT) << 4) | (val & IBS_OP_MAX_CNT);
+		val |= ibs_config.dispatched_ops ? IBS_OP_CNT_CTL : 0;
+		val |= IBS_OP_ENABLE;
+		ibs_state.ibs_op_ctl = val;
+		ibs_state.sample_size = IBS_OP_SIZE;
+		if (ibs_config.branch_target) {
+			ibs_state.branch_target = 1;
+			ibs_state.sample_size++;
 		}
-		if (ibs_caps & IBS_CAPS_OPCNT && ibs_config.dispatched_ops)
-			ibs_op_ctl |= IBS_OP_CNT_CTL;
-		ibs_op_ctl |= IBS_OP_ENABLE;
-		val = op_amd_randomize_ibs_op(ibs_op_ctl);
+		val = op_amd_randomize_ibs_op(ibs_state.ibs_op_ctl);
 		wrmsrl(MSR_AMD64_IBSOPCTL, val);
 	}
 }
@@ -281,29 +322,25 @@ static inline int eilvt_is_available(int offset)
 
 static inline int ibs_eilvt_valid(void)
 {
-	u64 val;
 	int offset;
+	u64 val;
 
 	rdmsrl(MSR_AMD64_IBSCTL, val);
+	offset = val & IBSCTL_LVT_OFFSET_MASK;
+
 	if (!(val & IBSCTL_LVT_OFFSET_VALID)) {
-		pr_err(FW_BUG "cpu %d, invalid IBS "
-		       "interrupt offset %d (MSR%08X=0x%016llx)",
-		       smp_processor_id(), offset,
-		       MSR_AMD64_IBSCTL, val);
+		pr_err(FW_BUG "cpu %d, invalid IBS interrupt offset %d (MSR%08X=0x%016llx)\n",
+		       smp_processor_id(), offset, MSR_AMD64_IBSCTL, val);
 		return 0;
 	}
 
-	offset = val & IBSCTL_LVT_OFFSET_MASK;
-
-	if (eilvt_is_available(offset))
-		return !0;
-
-	pr_err(FW_BUG "cpu %d, IBS interrupt offset %d "
-	       "not available (MSR%08X=0x%016llx)",
-	       smp_processor_id(), offset,
-	       MSR_AMD64_IBSCTL, val);
-
+	if (!eilvt_is_available(offset)) {
+		pr_err(FW_BUG "cpu %d, IBS interrupt offset %d not available (MSR%08X=0x%016llx)\n",
+		       smp_processor_id(), offset, MSR_AMD64_IBSCTL, val);
 		return 0;
+	}
+
+	return 1;
 }
 
 static inline int get_ibs_offset(void)
@@ -630,12 +667,11 @@ static int setup_ibs_files(struct super_block *sb, struct dentry *root)
 	/* model specific files */
 
 	/* setup some reasonable defaults */
+	memset(&ibs_config, 0, sizeof(ibs_config));
 	ibs_config.max_cnt_fetch = 250000;
-	ibs_config.fetch_enabled = 0;
 	ibs_config.max_cnt_op = 250000;
-	ibs_config.op_enabled = 0;
-	ibs_config.dispatched_ops = 0;
 
+	if (ibs_caps & IBS_CAPS_FETCHSAM) {
 		dir = oprofilefs_mkdir(sb, root, "ibs_fetch");
 		oprofilefs_create_ulong(sb, dir, "enable",
 					&ibs_config.fetch_enabled);
@@ -643,7 +679,9 @@ static int setup_ibs_files(struct super_block *sb, struct dentry *root)
 					&ibs_config.max_cnt_fetch);
 		oprofilefs_create_ulong(sb, dir, "rand_enable",
 					&ibs_config.rand_en);
+	}
 
+	if (ibs_caps & IBS_CAPS_OPSAM) {
 		dir = oprofilefs_mkdir(sb, root, "ibs_op");
 		oprofilefs_create_ulong(sb, dir, "enable",
 					&ibs_config.op_enabled);
@@ -652,6 +690,10 @@ static int setup_ibs_files(struct super_block *sb, struct dentry *root)
 		if (ibs_caps & IBS_CAPS_OPCNT)
 			oprofilefs_create_ulong(sb, dir, "dispatched_ops",
 						&ibs_config.dispatched_ops);
+		if (ibs_caps & IBS_CAPS_BRNTRGT)
+			oprofilefs_create_ulong(sb, dir, "branch_target",
+						&ibs_config.branch_target);
+	}
 
 	return 0;
 }

include/linux/interrupt.h

@@ -410,7 +410,7 @@ extern void open_softirq(int nr, void (*action)(struct softirq_action *));
 extern void softirq_init(void);
 static inline void __raise_softirq_irqoff(unsigned int nr)
 {
-	trace_softirq_raise((struct softirq_action *)(unsigned long)nr, NULL);
+	trace_softirq_raise(nr);
 	or_softirq_pending(1UL << nr);
 }
 

include/linux/ring_buffer.h

@@ -62,18 +62,6 @@ enum ring_buffer_type {
 unsigned ring_buffer_event_length(struct ring_buffer_event *event);
 void *ring_buffer_event_data(struct ring_buffer_event *event);
 
-/**
- * ring_buffer_event_time_delta - return the delta timestamp of the event
- * @event: the event to get the delta timestamp of
- *
- * The delta timestamp is the 27 bit timestamp since the last event.
- */
-static inline unsigned
-ring_buffer_event_time_delta(struct ring_buffer_event *event)
-{
-	return event->time_delta;
-}
-
 /*
  * ring_buffer_discard_commit will remove an event that has not
  *   ben committed yet. If this is used, then ring_buffer_unlock_commit

include/trace/events/irq.h

@@ -86,76 +86,62 @@ TRACE_EVENT(irq_handler_exit,
 
 DECLARE_EVENT_CLASS(softirq,
 
-	TP_PROTO(struct softirq_action *h, struct softirq_action *vec),
+	TP_PROTO(unsigned int vec_nr),
 
-	TP_ARGS(h, vec),
+	TP_ARGS(vec_nr),
 
 	TP_STRUCT__entry(
-		__field(	int,	vec			)
+		__field(	unsigned int,	vec	)
 	),
 
 	TP_fast_assign(
-		if (vec)
-			__entry->vec = (int)(h - vec);
-		else
-			__entry->vec = (int)(long)h;
+		__entry->vec = vec_nr;
 	),
 
-	TP_printk("vec=%d [action=%s]", __entry->vec,
+	TP_printk("vec=%u [action=%s]", __entry->vec,
 		  show_softirq_name(__entry->vec))
 );
 
 /**
  * softirq_entry - called immediately before the softirq handler
- * @h: pointer to struct softirq_action
- * @vec: pointer to first struct softirq_action in softirq_vec array
+ * @vec_nr:  softirq vector number
  *
- * The @h parameter, contains a pointer to the struct softirq_action
- * which has a pointer to the action handler that is called. By subtracting
- * the @vec pointer from the @h pointer, we can determine the softirq
- * number. Also, when used in combination with the softirq_exit tracepoint
- * we can determine the softirq latency.
+ * When used in combination with the softirq_exit tracepoint
+ * we can determine the softirq handler runtine.
  */
 DEFINE_EVENT(softirq, softirq_entry,
 
-	TP_PROTO(struct softirq_action *h, struct softirq_action *vec),
+	TP_PROTO(unsigned int vec_nr),
 
-	TP_ARGS(h, vec)
+	TP_ARGS(vec_nr)
 );
 
 /**
  * softirq_exit - called immediately after the softirq handler returns
- * @h: pointer to struct softirq_action
- * @vec: pointer to first struct softirq_action in softirq_vec array
+ * @vec_nr:  softirq vector number
  *
- * The @h parameter contains a pointer to the struct softirq_action
- * that has handled the softirq. By subtracting the @vec pointer from
- * the @h pointer, we can determine the softirq number. Also, when used in
- * combination with the softirq_entry tracepoint we can determine the softirq
- * latency.
+ * When used in combination with the softirq_entry tracepoint
+ * we can determine the softirq handler runtine.
  */
 DEFINE_EVENT(softirq, softirq_exit,
 
-	TP_PROTO(struct softirq_action *h, struct softirq_action *vec),
+	TP_PROTO(unsigned int vec_nr),
 
-	TP_ARGS(h, vec)
+	TP_ARGS(vec_nr)
 );
 
 /**
  * softirq_raise - called immediately when a softirq is raised
- * @h: pointer to struct softirq_action
- * @vec: pointer to first struct softirq_action in softirq_vec array
+ * @vec_nr:  softirq vector number
  *
- * The @h parameter contains a pointer to the softirq vector number which is
- * raised. @vec is NULL and it means @h includes vector number not
- * softirq_action. When used in combination with the softirq_entry tracepoint
- * we can determine the softirq raise latency.
+ * When used in combination with the softirq_entry tracepoint
+ * we can determine the softirq raise to run latency.
  */
 DEFINE_EVENT(softirq, softirq_raise,
 
-	TP_PROTO(struct softirq_action *h, struct softirq_action *vec),
+	TP_PROTO(unsigned int vec_nr),
 
-	TP_ARGS(h, vec)
+	TP_ARGS(vec_nr)
 );
 
 #endif /*  _TRACE_IRQ_H */

kernel/kprobes.c

@@ -74,7 +74,8 @@ static struct hlist_head kretprobe_inst_table[KPROBE_TABLE_SIZE];
 /* NOTE: change this value only with kprobe_mutex held */
 static bool kprobes_all_disarmed;
 
-static DEFINE_MUTEX(kprobe_mutex);	/* Protects kprobe_table */
+/* This protects kprobe_table and optimizing_list */
+static DEFINE_MUTEX(kprobe_mutex);
 static DEFINE_PER_CPU(struct kprobe *, kprobe_instance) = NULL;
 static struct {
 	spinlock_t lock ____cacheline_aligned_in_smp;
@@ -595,6 +596,7 @@ static __kprobes void try_to_optimize_kprobe(struct kprobe *p)
 }
 
 #ifdef CONFIG_SYSCTL
+/* This should be called with kprobe_mutex locked */
 static void __kprobes optimize_all_kprobes(void)
 {
 	struct hlist_head *head;
@@ -607,17 +609,16 @@ static void __kprobes optimize_all_kprobes(void)
 		return;
 
 	kprobes_allow_optimization = true;
-	mutex_lock(&text_mutex);
 	for (i = 0; i < KPROBE_TABLE_SIZE; i++) {
 		head = &kprobe_table[i];
 		hlist_for_each_entry_rcu(p, node, head, hlist)
 			if (!kprobe_disabled(p))
 				optimize_kprobe(p);
 	}
-	mutex_unlock(&text_mutex);
 	printk(KERN_INFO "Kprobes globally optimized\n");
 }
 
+/* This should be called with kprobe_mutex locked */
 static void __kprobes unoptimize_all_kprobes(void)
 {
 	struct hlist_head *head;

kernel/perf_event.c

@@ -417,8 +417,8 @@ event_filter_match(struct perf_event *event)
 	return event->cpu == -1 || event->cpu == smp_processor_id();
 }
 
-static int
-__event_sched_out(struct perf_event *event,
+static void
+event_sched_out(struct perf_event *event,
 		  struct perf_cpu_context *cpuctx,
 		  struct perf_event_context *ctx)
 {
@@ -437,13 +437,14 @@ __event_sched_out(struct perf_event *event,
 	}
 
 	if (event->state != PERF_EVENT_STATE_ACTIVE)
-		return 0;
+		return;
 
 	event->state = PERF_EVENT_STATE_INACTIVE;
 	if (event->pending_disable) {
 		event->pending_disable = 0;
 		event->state = PERF_EVENT_STATE_OFF;
 	}
+	event->tstamp_stopped = ctx->time;
 	event->pmu->del(event, 0);
 	event->oncpu = -1;
 
@@ -452,19 +453,6 @@ __event_sched_out(struct perf_event *event,
 	ctx->nr_active--;
 	if (event->attr.exclusive || !cpuctx->active_oncpu)
 		cpuctx->exclusive = 0;
-	return 1;
-}
-
-static void
-event_sched_out(struct perf_event *event,
-		  struct perf_cpu_context *cpuctx,
-		  struct perf_event_context *ctx)
-{
-	int ret;
-
-	ret = __event_sched_out(event, cpuctx, ctx);
-	if (ret)
-		event->tstamp_stopped = ctx->time;
 }
 
 static void
@@ -664,7 +652,7 @@ void perf_event_disable(struct perf_event *event)
 }
 
 static int
-__event_sched_in(struct perf_event *event,
+event_sched_in(struct perf_event *event,
 		 struct perf_cpu_context *cpuctx,
 		 struct perf_event_context *ctx)
 {
@@ -684,6 +672,8 @@ __event_sched_in(struct perf_event *event,
 		return -EAGAIN;
 	}
 
+	event->tstamp_running += ctx->time - event->tstamp_stopped;
+
 	if (!is_software_event(event))
 		cpuctx->active_oncpu++;
 	ctx->nr_active++;
@@ -694,35 +684,6 @@ __event_sched_in(struct perf_event *event,
 	return 0;
 }
 
-static inline int
-event_sched_in(struct perf_event *event,
-		 struct perf_cpu_context *cpuctx,
-		 struct perf_event_context *ctx)
-{
-	int ret = __event_sched_in(event, cpuctx, ctx);
-	if (ret)
-		return ret;
-	event->tstamp_running += ctx->time - event->tstamp_stopped;
-	return 0;
-}
-
-static void
-group_commit_event_sched_in(struct perf_event *group_event,
-	       struct perf_cpu_context *cpuctx,
-	       struct perf_event_context *ctx)
-{
-	struct perf_event *event;
-	u64 now = ctx->time;
-
-	group_event->tstamp_running += now - group_event->tstamp_stopped;
-	/*
-	 * Schedule in siblings as one group (if any):
-	 */
-	list_for_each_entry(event, &group_event->sibling_list, group_entry) {
-		event->tstamp_running += now - event->tstamp_stopped;
-	}
-}
-
 static int
 group_sched_in(struct perf_event *group_event,
 	       struct perf_cpu_context *cpuctx,
@@ -730,19 +691,15 @@ group_sched_in(struct perf_event *group_event,
 {
 	struct perf_event *event, *partial_group = NULL;
 	struct pmu *pmu = group_event->pmu;
+	u64 now = ctx->time;
+	bool simulate = false;
 
 	if (group_event->state == PERF_EVENT_STATE_OFF)
 		return 0;
 
 	pmu->start_txn(pmu);
 
-	/*
-	 * use __event_sched_in() to delay updating tstamp_running
-	 * until the transaction is committed. In case of failure
-	 * we will keep an unmodified tstamp_running which is a
-	 * requirement to get correct timing information
-	 */
-	if (__event_sched_in(group_event, cpuctx, ctx)) {
+	if (event_sched_in(group_event, cpuctx, ctx)) {
 		pmu->cancel_txn(pmu);
 		return -EAGAIN;
 	}
@@ -751,31 +708,42 @@ group_sched_in(struct perf_event *group_event,
 	 * Schedule in siblings as one group (if any):
 	 */
 	list_for_each_entry(event, &group_event->sibling_list, group_entry) {
-		if (__event_sched_in(event, cpuctx, ctx)) {
+		if (event_sched_in(event, cpuctx, ctx)) {
 			partial_group = event;
 			goto group_error;
 		}
 	}
 
-	if (!pmu->commit_txn(pmu)) {
-		/* commit tstamp_running */
-		group_commit_event_sched_in(group_event, cpuctx, ctx);
+	if (!pmu->commit_txn(pmu))
 		return 0;
-	}
+
 group_error:
 	/*
 	 * Groups can be scheduled in as one unit only, so undo any
 	 * partial group before returning:
+	 * The events up to the failed event are scheduled out normally,
+	 * tstamp_stopped will be updated.
 	 *
-	 * use __event_sched_out() to avoid updating tstamp_stopped
-	 * because the event never actually ran
+	 * The failed events and the remaining siblings need to have
+	 * their timings updated as if they had gone thru event_sched_in()
+	 * and event_sched_out(). This is required to get consistent timings
+	 * across the group. This also takes care of the case where the group
+	 * could never be scheduled by ensuring tstamp_stopped is set to mark
+	 * the time the event was actually stopped, such that time delta
+	 * calculation in update_event_times() is correct.
 	 */
 	list_for_each_entry(event, &group_event->sibling_list, group_entry) {
 		if (event == partial_group)
-			break;
-		__event_sched_out(event, cpuctx, ctx);
+			simulate = true;
+
+		if (simulate) {
+			event->tstamp_running += now - event->tstamp_stopped;
+			event->tstamp_stopped = now;
+		} else {
+			event_sched_out(event, cpuctx, ctx);
+		}
 	}
-	__event_sched_out(group_event, cpuctx, ctx);
+	event_sched_out(group_event, cpuctx, ctx);
 
 	pmu->cancel_txn(pmu);
 

kernel/softirq.c

@@ -229,18 +229,20 @@ asmlinkage void __do_softirq(void)
 
 	do {
 		if (pending & 1) {
+			unsigned int vec_nr = h - softirq_vec;
 			int prev_count = preempt_count();
-			kstat_incr_softirqs_this_cpu(h - softirq_vec);
 
-			trace_softirq_entry(h, softirq_vec);
+			kstat_incr_softirqs_this_cpu(vec_nr);
+
+			trace_softirq_entry(vec_nr);
 			h->action(h);
-			trace_softirq_exit(h, softirq_vec);
+			trace_softirq_exit(vec_nr);
 			if (unlikely(prev_count != preempt_count())) {
-				printk(KERN_ERR "huh, entered softirq %td %s %p"
+				printk(KERN_ERR "huh, entered softirq %u %s %p"
 				       "with preempt_count %08x,"
-				       " exited with %08x?\n", h - softirq_vec,
-				       softirq_to_name[h - softirq_vec],
-				       h->action, prev_count, preempt_count());
+				       " exited with %08x?\n", vec_nr,
+				       softirq_to_name[vec_nr], h->action,
+				       prev_count, preempt_count());
 				preempt_count() = prev_count;
 			}
 

kernel/trace/ring_buffer.c

@@ -224,6 +224,9 @@ enum {
 	RB_LEN_TIME_STAMP = 16,
 };
 
+#define skip_time_extend(event) \
+	((struct ring_buffer_event *)((char *)event + RB_LEN_TIME_EXTEND))
+
 static inline int rb_null_event(struct ring_buffer_event *event)
 {
 	return event->type_len == RINGBUF_TYPE_PADDING && !event->time_delta;
@@ -248,8 +251,12 @@ rb_event_data_length(struct ring_buffer_event *event)
 	return length + RB_EVNT_HDR_SIZE;
 }
 
-/* inline for ring buffer fast paths */
-static unsigned
+/*
+ * Return the length of the given event. Will return
+ * the length of the time extend if the event is a
+ * time extend.
+ */
+static inline unsigned
 rb_event_length(struct ring_buffer_event *event)
 {
 	switch (event->type_len) {
@@ -274,13 +281,41 @@ rb_event_length(struct ring_buffer_event *event)
 	return 0;
 }
 
+/*
+ * Return total length of time extend and data,
+ *   or just the event length for all other events.
+ */
+static inline unsigned
+rb_event_ts_length(struct ring_buffer_event *event)
+{
+	unsigned len = 0;
+
+	if (event->type_len == RINGBUF_TYPE_TIME_EXTEND) {
+		/* time extends include the data event after it */
+		len = RB_LEN_TIME_EXTEND;
+		event = skip_time_extend(event);
+	}
+	return len + rb_event_length(event);
+}
+
 /**
  * ring_buffer_event_length - return the length of the event
  * @event: the event to get the length of
+ *
+ * Returns the size of the data load of a data event.
+ * If the event is something other than a data event, it
+ * returns the size of the event itself. With the exception
+ * of a TIME EXTEND, where it still returns the size of the
+ * data load of the data event after it.
  */
 unsigned ring_buffer_event_length(struct ring_buffer_event *event)
 {
-	unsigned length = rb_event_length(event);
+	unsigned length;
+
+	if (event->type_len == RINGBUF_TYPE_TIME_EXTEND)
+		event = skip_time_extend(event);
+
+	length = rb_event_length(event);
 	if (event->type_len > RINGBUF_TYPE_DATA_TYPE_LEN_MAX)
 		return length;
 	length -= RB_EVNT_HDR_SIZE;
@@ -294,6 +329,8 @@ EXPORT_SYMBOL_GPL(ring_buffer_event_length);
 static void *
 rb_event_data(struct ring_buffer_event *event)
 {
+	if (event->type_len == RINGBUF_TYPE_TIME_EXTEND)
+		event = skip_time_extend(event);
 	BUG_ON(event->type_len > RINGBUF_TYPE_DATA_TYPE_LEN_MAX);
 	/* If length is in len field, then array[0] has the data */
 	if (event->type_len)
@@ -404,9 +441,6 @@ static inline int test_time_stamp(u64 delta)
 /* Max payload is BUF_PAGE_SIZE - header (8bytes) */
 #define BUF_MAX_DATA_SIZE (BUF_PAGE_SIZE - (sizeof(u32) * 2))
 
-/* Max number of timestamps that can fit on a page */
-#define RB_TIMESTAMPS_PER_PAGE	(BUF_PAGE_SIZE / RB_LEN_TIME_EXTEND)
-
 int ring_buffer_print_page_header(struct trace_seq *s)
 {
 	struct buffer_data_page field;
@@ -1546,6 +1580,25 @@ static void rb_inc_iter(struct ring_buffer_iter *iter)
 	iter->head = 0;
 }
 
+/* Slow path, do not inline */
+static noinline struct ring_buffer_event *
+rb_add_time_stamp(struct ring_buffer_event *event, u64 delta)
+{
+	event->type_len = RINGBUF_TYPE_TIME_EXTEND;
+
+	/* Not the first event on the page? */
+	if (rb_event_index(event)) {
+		event->time_delta = delta & TS_MASK;
+		event->array[0] = delta >> TS_SHIFT;
+	} else {
+		/* nope, just zero it */
+		event->time_delta = 0;
+		event->array[0] = 0;
+	}
+
+	return skip_time_extend(event);
+}
+
 /**
  * ring_buffer_update_event - update event type and data
  * @event: the even to update
@@ -1558,28 +1611,31 @@ static void rb_inc_iter(struct ring_buffer_iter *iter)
  * data field.
  */
 static void
-rb_update_event(struct ring_buffer_event *event,
-			 unsigned type, unsigned length)
+rb_update_event(struct ring_buffer_per_cpu *cpu_buffer,
+		struct ring_buffer_event *event, unsigned length,
+		int add_timestamp, u64 delta)
 {
-	event->type_len = type;
-
-	switch (type) {
+	/* Only a commit updates the timestamp */
+	if (unlikely(!rb_event_is_commit(cpu_buffer, event)))
+		delta = 0;
 
-	case RINGBUF_TYPE_PADDING:
-	case RINGBUF_TYPE_TIME_EXTEND:
-	case RINGBUF_TYPE_TIME_STAMP:
-		break;
+	/*
+	 * If we need to add a timestamp, then we
+	 * add it to the start of the resevered space.
+	 */
+	if (unlikely(add_timestamp)) {
+		event = rb_add_time_stamp(event, delta);
+		length -= RB_LEN_TIME_EXTEND;
+		delta = 0;
+	}
 
-	case 0:
+	event->time_delta = delta;
 	length -= RB_EVNT_HDR_SIZE;
-		if (length > RB_MAX_SMALL_DATA || RB_FORCE_8BYTE_ALIGNMENT)
+	if (length > RB_MAX_SMALL_DATA || RB_FORCE_8BYTE_ALIGNMENT) {
+		event->type_len = 0;
 		event->array[0] = length;
-		else
+	} else
 		event->type_len = DIV_ROUND_UP(length, RB_ALIGNMENT);
-		break;
-	default:
-		BUG();
-	}
 }
 
 /*
@@ -1823,10 +1879,13 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	local_sub(length, &tail_page->write);
 }
 
-static struct ring_buffer_event *
+/*
+ * This is the slow path, force gcc not to inline it.
+ */
+static noinline struct ring_buffer_event *
 rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	     unsigned long length, unsigned long tail,
-	     struct buffer_page *tail_page, u64 *ts)
+	     struct buffer_page *tail_page, u64 ts)
 {
 	struct buffer_page *commit_page = cpu_buffer->commit_page;
 	struct ring_buffer *buffer = cpu_buffer->buffer;
@@ -1909,8 +1968,8 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 		 * Nested commits always have zero deltas, so
 		 * just reread the time stamp
 		 */
-		*ts = rb_time_stamp(buffer);
-		next_page->page->time_stamp = *ts;
+		ts = rb_time_stamp(buffer);
+		next_page->page->time_stamp = ts;
 	}
 
  out_again:
@@ -1929,12 +1988,21 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 
 static struct ring_buffer_event *
 __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
-		  unsigned type, unsigned long length, u64 *ts)
+		  unsigned long length, u64 ts,
+		  u64 delta, int add_timestamp)
 {
 	struct buffer_page *tail_page;
 	struct ring_buffer_event *event;
 	unsigned long tail, write;
 
+	/*
+	 * If the time delta since the last event is too big to
+	 * hold in the time field of the event, then we append a
+	 * TIME EXTEND event ahead of the data event.
+	 */
+	if (unlikely(add_timestamp))
+		length += RB_LEN_TIME_EXTEND;
+
 	tail_page = cpu_buffer->tail_page;
 	write = local_add_return(length, &tail_page->write);
 
@@ -1943,7 +2011,7 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
 	tail = write - length;
 
 	/* See if we shot pass the end of this buffer page */
-	if (write > BUF_PAGE_SIZE)
+	if (unlikely(write > BUF_PAGE_SIZE))
 		return rb_move_tail(cpu_buffer, length, tail,
 				    tail_page, ts);
 
@@ -1951,10 +2019,8 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
 
 	event = __rb_page_index(tail_page, tail);
 	kmemcheck_annotate_bitfield(event, bitfield);
-	rb_update_event(event, type, length);
+	rb_update_event(cpu_buffer, event, length, add_timestamp, delta);
 
-	/* The passed in type is zero for DATA */
-	if (likely(!type))
 	local_inc(&tail_page->entries);
 
 	/*
@@ -1962,7 +2028,7 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
 	 * its timestamp.
 	 */
 	if (!tail)
-		tail_page->page->time_stamp = *ts;
+		tail_page->page->time_stamp = ts;
 
 	return event;
 }
@@ -1977,7 +2043,7 @@ rb_try_to_discard(struct ring_buffer_per_cpu *cpu_buffer,
 	unsigned long addr;
 
 	new_index = rb_event_index(event);
-	old_index = new_index + rb_event_length(event);
+	old_index = new_index + rb_event_ts_length(event);
 	addr = (unsigned long)event;
 	addr &= PAGE_MASK;
 
@@ -2003,76 +2069,13 @@ rb_try_to_discard(struct ring_buffer_per_cpu *cpu_buffer,
 	return 0;
 }
 
-static int
-rb_add_time_stamp(struct ring_buffer_per_cpu *cpu_buffer,
-		  u64 *ts, u64 *delta)
-{
-	struct ring_buffer_event *event;
-	int ret;
-
-	WARN_ONCE(*delta > (1ULL << 59),
-		  KERN_WARNING "Delta way too big! %llu ts=%llu write stamp = %llu\n",
-		  (unsigned long long)*delta,
-		  (unsigned long long)*ts,
-		  (unsigned long long)cpu_buffer->write_stamp);
-
-	/*
-	 * The delta is too big, we to add a
-	 * new timestamp.
-	 */
-	event = __rb_reserve_next(cpu_buffer,
-				  RINGBUF_TYPE_TIME_EXTEND,
-				  RB_LEN_TIME_EXTEND,
-				  ts);
-	if (!event)
-		return -EBUSY;
-
-	if (PTR_ERR(event) == -EAGAIN)
-		return -EAGAIN;
-
-	/* Only a commited time event can update the write stamp */
-	if (rb_event_is_commit(cpu_buffer, event)) {
-		/*
-		 * If this is the first on the page, then it was
-		 * updated with the page itself. Try to discard it
-		 * and if we can't just make it zero.
-		 */
-		if (rb_event_index(event)) {
-			event->time_delta = *delta & TS_MASK;
-			event->array[0] = *delta >> TS_SHIFT;
-		} else {
-			/* try to discard, since we do not need this */
-			if (!rb_try_to_discard(cpu_buffer, event)) {
-				/* nope, just zero it */
-				event->time_delta = 0;
-				event->array[0] = 0;
-			}
-		}
-		cpu_buffer->write_stamp = *ts;
-		/* let the caller know this was the commit */
-		ret = 1;
-	} else {
-		/* Try to discard the event */
-		if (!rb_try_to_discard(cpu_buffer, event)) {
-			/* Darn, this is just wasted space */
-			event->time_delta = 0;
-			event->array[0] = 0;
-		}
-		ret = 0;
-	}
-
-	*delta = 0;
-
-	return ret;
-}
-
 static void rb_start_commit(struct ring_buffer_per_cpu *cpu_buffer)
 {
 	local_inc(&cpu_buffer->committing);
 	local_inc(&cpu_buffer->commits);
 }
 
-static void rb_end_commit(struct ring_buffer_per_cpu *cpu_buffer)
+static inline void rb_end_commit(struct ring_buffer_per_cpu *cpu_buffer)
 {
 	unsigned long commits;
 
@@ -2110,9 +2113,10 @@ rb_reserve_next_event(struct ring_buffer *buffer,
 		      unsigned long length)
 {
 	struct ring_buffer_event *event;
-	u64 ts, delta = 0;
-	int commit = 0;
+	u64 ts, delta;
 	int nr_loops = 0;
+	int add_timestamp;
+	u64 diff;
 
 	rb_start_commit(cpu_buffer);
 
@@ -2133,6 +2137,9 @@ rb_reserve_next_event(struct ring_buffer *buffer,
 
 	length = rb_calculate_event_length(length);
  again:
+	add_timestamp = 0;
+	delta = 0;
+
 	/*
 	 * We allow for interrupts to reenter here and do a trace.
 	 * If one does, it will cause this original code to loop
@@ -2146,56 +2153,32 @@ rb_reserve_next_event(struct ring_buffer *buffer,
 		goto out_fail;
 
 	ts = rb_time_stamp(cpu_buffer->buffer);
-
-	/*
-	 * Only the first commit can update the timestamp.
-	 * Yes there is a race here. If an interrupt comes in
-	 * just after the conditional and it traces too, then it
-	 * will also check the deltas. More than one timestamp may
-	 * also be made. But only the entry that did the actual
-	 * commit will be something other than zero.
-	 */
-	if (likely(cpu_buffer->tail_page == cpu_buffer->commit_page &&
-		   rb_page_write(cpu_buffer->tail_page) ==
-		   rb_commit_index(cpu_buffer))) {
-		u64 diff;
-
 	diff = ts - cpu_buffer->write_stamp;
 
 	/* make sure this diff is calculated here */
 	barrier();
 
 	/* Did the write stamp get updated already? */
-		if (unlikely(ts < cpu_buffer->write_stamp))
-			goto get_event;
-
+	if (likely(ts >= cpu_buffer->write_stamp)) {
 		delta = diff;
 		if (unlikely(test_time_stamp(delta))) {
-
-			commit = rb_add_time_stamp(cpu_buffer, &ts, &delta);
-			if (commit == -EBUSY)
-				goto out_fail;
-
-			if (commit == -EAGAIN)
-				goto again;
-
-			RB_WARN_ON(cpu_buffer, commit < 0);
+			WARN_ONCE(delta > (1ULL << 59),
+				  KERN_WARNING "Delta way too big! %llu ts=%llu write stamp = %llu\n",
+				  (unsigned long long)delta,
+				  (unsigned long long)ts,
+				  (unsigned long long)cpu_buffer->write_stamp);
+			add_timestamp = 1;
 		}
 	}
 
- get_event:
-	event = __rb_reserve_next(cpu_buffer, 0, length, &ts);
+	event = __rb_reserve_next(cpu_buffer, length, ts,
+				  delta, add_timestamp);
 	if (unlikely(PTR_ERR(event) == -EAGAIN))
 		goto again;
 
 	if (!event)
 		goto out_fail;
 
-	if (!rb_event_is_commit(cpu_buffer, event))
-		delta = 0;
-
-	event->time_delta = delta;
-
 	return event;
 
  out_fail:
@@ -2207,13 +2190,9 @@ rb_reserve_next_event(struct ring_buffer *buffer,
 
 #define TRACE_RECURSIVE_DEPTH 16
 
-static int trace_recursive_lock(void)
+/* Keep this code out of the fast path cache */
+static noinline void trace_recursive_fail(void)
 {
-	current->trace_recursion++;
-
-	if (likely(current->trace_recursion < TRACE_RECURSIVE_DEPTH))
-		return 0;
-
 	/* Disable all tracing before we do anything else */
 	tracing_off_permanent();
 
@@ -2225,10 +2204,21 @@ static int trace_recursive_lock(void)
 		    in_nmi());
 
 	WARN_ON_ONCE(1);
+}
+
+static inline int trace_recursive_lock(void)
+{
+	current->trace_recursion++;
+
+	if (likely(current->trace_recursion < TRACE_RECURSIVE_DEPTH))
+		return 0;
+
+	trace_recursive_fail();
+
 	return -1;
 }
 
-static void trace_recursive_unlock(void)
+static inline void trace_recursive_unlock(void)
 {
 	WARN_ON_ONCE(!current->trace_recursion);
 
@@ -2308,12 +2298,28 @@ static void
 rb_update_write_stamp(struct ring_buffer_per_cpu *cpu_buffer,
 		      struct ring_buffer_event *event)
 {
+	u64 delta;
+
 	/*
 	 * The event first in the commit queue updates the
 	 * time stamp.
 	 */
-	if (rb_event_is_commit(cpu_buffer, event))
+	if (rb_event_is_commit(cpu_buffer, event)) {
+		/*
+		 * A commit event that is first on a page
+		 * updates the write timestamp with the page stamp
+		 */
+		if (!rb_event_index(event))
+			cpu_buffer->write_stamp =
+				cpu_buffer->commit_page->page->time_stamp;
+		else if (event->type_len == RINGBUF_TYPE_TIME_EXTEND) {
+			delta = event->array[0];
+			delta <<= TS_SHIFT;
+			delta += event->time_delta;
+			cpu_buffer->write_stamp += delta;
+		} else
 			cpu_buffer->write_stamp += event->time_delta;
+	}
 }
 
 static void rb_commit(struct ring_buffer_per_cpu *cpu_buffer,
@@ -2353,6 +2359,9 @@ EXPORT_SYMBOL_GPL(ring_buffer_unlock_commit);
 
 static inline void rb_event_discard(struct ring_buffer_event *event)
 {
+	if (event->type_len == RINGBUF_TYPE_TIME_EXTEND)
+		event = skip_time_extend(event);
+
 	/* array[0] holds the actual length for the discarded event */
 	event->array[0] = rb_event_data_length(event) - RB_EVNT_HDR_SIZE;
 	event->type_len = RINGBUF_TYPE_PADDING;
@@ -3049,12 +3058,12 @@ rb_buffer_peek(struct ring_buffer_per_cpu *cpu_buffer, u64 *ts,
 
  again:
 	/*
-	 * We repeat when a timestamp is encountered. It is possible
-	 * to get multiple timestamps from an interrupt entering just
-	 * as one timestamp is about to be written, or from discarded
-	 * commits. The most that we can have is the number on a single page.
+	 * We repeat when a time extend is encountered.
+	 * Since the time extend is always attached to a data event,
+	 * we should never loop more than once.
+	 * (We never hit the following condition more than twice).
 	 */
-	if (RB_WARN_ON(cpu_buffer, ++nr_loops > RB_TIMESTAMPS_PER_PAGE))
+	if (RB_WARN_ON(cpu_buffer, ++nr_loops > 2))
 		return NULL;
 
 	reader = rb_get_reader_page(cpu_buffer);
@@ -3130,14 +3139,12 @@ rb_iter_peek(struct ring_buffer_iter *iter, u64 *ts)
 		return NULL;
 
 	/*
-	 * We repeat when a timestamp is encountered.
-	 * We can get multiple timestamps by nested interrupts or also
-	 * if filtering is on (discarding commits). Since discarding
-	 * commits can be frequent we can get a lot of timestamps.
-	 * But we limit them by not adding timestamps if they begin
-	 * at the start of a page.
+	 * We repeat when a time extend is encountered.
+	 * Since the time extend is always attached to a data event,
+	 * we should never loop more than once.
+	 * (We never hit the following condition more than twice).
 	 */
-	if (RB_WARN_ON(cpu_buffer, ++nr_loops > RB_TIMESTAMPS_PER_PAGE))
+	if (RB_WARN_ON(cpu_buffer, ++nr_loops > 2))
 		return NULL;
 
 	if (rb_per_cpu_empty(cpu_buffer))
@@ -3835,7 +3842,8 @@ int ring_buffer_read_page(struct ring_buffer *buffer,
 		if (len > (commit - read))
 			len = (commit - read);
 
-		size = rb_event_length(event);
+		/* Always keep the time extend and data together */
+		size = rb_event_ts_length(event);
 
 		if (len < size)
 			goto out_unlock;
@@ -3857,7 +3865,8 @@ int ring_buffer_read_page(struct ring_buffer *buffer,
 				break;
 
 			event = rb_reader_event(cpu_buffer);
-			size = rb_event_length(event);
+			/* Always keep the time extend and data together */
+			size = rb_event_ts_length(event);
 		} while (len > size);
 
 		/* update bpage */

kernel/trace/trace.c

@@ -3996,13 +3996,9 @@ static void tracing_init_debugfs_percpu(long cpu)
 {
 	struct dentry *d_percpu = tracing_dentry_percpu();
 	struct dentry *d_cpu;
-	/* strlen(cpu) + MAX(log10(cpu)) + '\0' */
-	char cpu_dir[7];
+	char cpu_dir[30]; /* 30 characters should be more than enough */
 
-	if (cpu > 999 || cpu < 0)
-		return;
-
-	sprintf(cpu_dir, "cpu%ld", cpu);
+	snprintf(cpu_dir, 30, "cpu%ld", cpu);
 	d_cpu = debugfs_create_dir(cpu_dir, d_percpu);
 	if (!d_cpu) {
 		pr_warning("Could not create debugfs '%s' entry\n", cpu_dir);

tools/perf/Documentation/perf-list.txt

@@ -15,6 +15,23 @@ DESCRIPTION
 This command displays the symbolic event types which can be selected in the
 various perf commands with the -e option.
 
+EVENT MODIFIERS
+---------------
+
+Events can optionally have a modifer by appending a colon and one or
+more modifiers.  Modifiers allow the user to restrict when events are
+counted with 'u' for user-space, 'k' for kernel, 'h' for hypervisor.
+
+The 'p' modifier can be used for specifying how precise the instruction
+address should be. The 'p' modifier is currently only implemented for
+Intel PEBS and can be specified multiple times:
+  0 - SAMPLE_IP can have arbitrary skid
+  1 - SAMPLE_IP must have constant skid
+  2 - SAMPLE_IP requested to have 0 skid
+  3 - SAMPLE_IP must have 0 skid
+
+The PEBS implementation now supports up to 2.
+
 RAW HARDWARE EVENT DESCRIPTOR
 -----------------------------
 Even when an event is not available in a symbolic form within perf right now,

tools/perf/Documentation/perf-probe.txt

@@ -16,7 +16,9 @@ or
 or
 'perf probe' --list
 or
-'perf probe' --line='FUNC[:RLN[+NUM|:RLN2]]|SRC:ALN[+NUM|:ALN2]'
+'perf probe' [options] --line='FUNC[:RLN[+NUM|:RLN2]]|SRC:ALN[+NUM|:ALN2]'
+or
+'perf probe' [options] --vars='PROBEPOINT'
 
 DESCRIPTION
 -----------
@@ -31,6 +33,11 @@ OPTIONS
 --vmlinux=PATH::
 	Specify vmlinux path which has debuginfo (Dwarf binary).
 
+-m::
+--module=MODNAME::
+	Specify module name in which perf-probe searches probe points
+	or lines.
+
 -s::
 --source=PATH::
 	Specify path to kernel source.
@@ -57,6 +64,15 @@ OPTIONS
 	Show source code lines which can be probed. This needs an argument
 	which specifies a range of the source code. (see LINE SYNTAX for detail)
 
+-V::
+--vars=::
+	Show available local variables at given probe point. The argument
+	syntax is same as PROBE SYNTAX, but NO ARGs.
+
+--externs::
+	(Only for --vars) Show external defined variables in addition to local
+	variables.
+
 -f::
 --force::
 	Forcibly add events with existing name.

tools/perf/Documentation/perf-record.txt

@@ -83,6 +83,10 @@ OPTIONS
 --call-graph::
 	Do call-graph (stack chain/backtrace) recording.
 
+-q::
+--quiet::
+	Don't print any message, useful for scripting.
+
 -v::
 --verbose::
 	Be more verbose (show counter open errors, etc).

tools/perf/builtin-probe.c

@@ -50,14 +50,17 @@ static struct {
 	bool list_events;
 	bool force_add;
 	bool show_lines;
+	bool show_vars;
+	bool show_ext_vars;
+	bool mod_events;
 	int nevents;
 	struct perf_probe_event events[MAX_PROBES];
 	struct strlist *dellist;
 	struct line_range line_range;
+	const char *target_module;
 	int max_probe_points;
 } params;
 
-
 /* Parse an event definition. Note that any error must die. */
 static int parse_probe_event(const char *str)
 {
@@ -92,6 +95,7 @@ static int parse_probe_event_argv(int argc, const char **argv)
 	len = 0;
 	for (i = 0; i < argc; i++)
 		len += sprintf(&buf[len], "%s ", argv[i]);
+	params.mod_events = true;
 	ret = parse_probe_event(buf);
 	free(buf);
 	return ret;
@@ -100,9 +104,10 @@ static int parse_probe_event_argv(int argc, const char **argv)
 static int opt_add_probe_event(const struct option *opt __used,
 			      const char *str, int unset __used)
 {
-	if (str)
+	if (str) {
+		params.mod_events = true;
 		return parse_probe_event(str);
-	else
+	} else
 		return 0;
 }
 
@@ -110,6 +115,7 @@ static int opt_del_probe_event(const struct option *opt __used,
 			       const char *str, int unset __used)
 {
 	if (str) {
+		params.mod_events = true;
 		if (!params.dellist)
 			params.dellist = strlist__new(true, NULL);
 		strlist__add(params.dellist, str);
@@ -130,6 +136,25 @@ static int opt_show_lines(const struct option *opt __used,
 
 	return ret;
 }
+
+static int opt_show_vars(const struct option *opt __used,
+			 const char *str, int unset __used)
+{
+	struct perf_probe_event *pev = &params.events[params.nevents];
+	int ret;
+
+	if (!str)
+		return 0;
+
+	ret = parse_probe_event(str);
+	if (!ret && pev->nargs != 0) {
+		pr_err("  Error: '--vars' doesn't accept arguments.\n");
+		return -EINVAL;
+	}
+	params.show_vars = true;
+
+	return ret;
+}
 #endif
 
 static const char * const probe_usage[] = {
@@ -138,7 +163,8 @@ static const char * const probe_usage[] = {
 	"perf probe [<options>] --del '[GROUP:]EVENT' ...",
 	"perf probe --list",
 #ifdef DWARF_SUPPORT
-	"perf probe --line 'LINEDESC'",
+	"perf probe [<options>] --line 'LINEDESC'",
+	"perf probe [<options>] --vars 'PROBEPOINT'",
 #endif
 	NULL
 };
@@ -180,10 +206,17 @@ static const struct option options[] = {
 	OPT_CALLBACK('L', "line", NULL,
 		     "FUNC[:RLN[+NUM|-RLN2]]|SRC:ALN[+NUM|-ALN2]",
 		     "Show source code lines.", opt_show_lines),
+	OPT_CALLBACK('V', "vars", NULL,
+		     "FUNC[@SRC][+OFF|%return|:RL|;PT]|SRC:AL|SRC;PT",
+		     "Show accessible variables on PROBEDEF", opt_show_vars),
+	OPT_BOOLEAN('\0', "externs", &params.show_ext_vars,
+		    "Show external variables too (with --vars only)"),
 	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
 		   "file", "vmlinux pathname"),
 	OPT_STRING('s', "source", &symbol_conf.source_prefix,
 		   "directory", "path to kernel source"),
+	OPT_STRING('m', "module", &params.target_module,
+		   "modname", "target module name"),
 #endif
 	OPT__DRY_RUN(&probe_event_dry_run),
 	OPT_INTEGER('\0', "max-probes", &params.max_probe_points,
@@ -217,7 +250,7 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used)
 		usage_with_options(probe_usage, options);
 
 	if (params.list_events) {
-		if (params.nevents != 0 || params.dellist) {
+		if (params.mod_events) {
 			pr_err("  Error: Don't use --list with --add/--del.\n");
 			usage_with_options(probe_usage, options);
 		}
@@ -225,6 +258,10 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used)
 			pr_err("  Error: Don't use --list with --line.\n");
 			usage_with_options(probe_usage, options);
 		}
+		if (params.show_vars) {
+			pr_err(" Error: Don't use --list with --vars.\n");
+			usage_with_options(probe_usage, options);
+		}
 		ret = show_perf_probe_events();
 		if (ret < 0)
 			pr_err("  Error: Failed to show event list. (%d)\n",
@@ -234,17 +271,35 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used)
 
 #ifdef DWARF_SUPPORT
 	if (params.show_lines) {
-		if (params.nevents != 0 || params.dellist) {
-			pr_warning("  Error: Don't use --line with"
+		if (params.mod_events) {
+			pr_err("  Error: Don't use --line with"
 			       " --add/--del.\n");
 			usage_with_options(probe_usage, options);
 		}
+		if (params.show_vars) {
+			pr_err(" Error: Don't use --line with --vars.\n");
+			usage_with_options(probe_usage, options);
+		}
 
-		ret = show_line_range(&params.line_range);
+		ret = show_line_range(&params.line_range, params.target_module);
 		if (ret < 0)
 			pr_err("  Error: Failed to show lines. (%d)\n", ret);
 		return ret;
 	}
+	if (params.show_vars) {
+		if (params.mod_events) {
+			pr_err("  Error: Don't use --vars with"
+			       " --add/--del.\n");
+			usage_with_options(probe_usage, options);
+		}
+		ret = show_available_vars(params.events, params.nevents,
+					  params.max_probe_points,
+					  params.target_module,
+					  params.show_ext_vars);
+		if (ret < 0)
+			pr_err("  Error: Failed to show vars. (%d)\n", ret);
+		return ret;
+	}
 #endif
 
 	if (params.dellist) {
@@ -258,8 +313,9 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used)
 
 	if (params.nevents) {
 		ret = add_perf_probe_events(params.events, params.nevents,
-					    params.force_add,
-					    params.max_probe_points);
+					    params.max_probe_points,
+					    params.target_module,
+					    params.force_add);
 		if (ret < 0) {
 			pr_err("  Error: Failed to add events. (%d)\n", ret);
 			return ret;

tools/perf/builtin-record.c

@@ -353,7 +353,7 @@ static void create_counter(int counter, int cpu)
 		}
 
 		if (read(fd[nr_cpu][counter][thread_index], &read_data, sizeof(read_data)) == -1) {
-			perror("Unable to read perf file descriptor\n");
+			perror("Unable to read perf file descriptor");
 			exit(-1);
 		}
 
@@ -626,7 +626,7 @@ static int __cmd_record(int argc, const char **argv)
 
 	nr_cpus = read_cpu_map(cpu_list);
 	if (nr_cpus < 1) {
-		perror("failed to collect number of CPUs\n");
+		perror("failed to collect number of CPUs");
 		return -1;
 	}
 
@@ -761,6 +761,9 @@ static int __cmd_record(int argc, const char **argv)
 		}
 	}
 
+	if (quiet)
+		return 0;
+
 	fprintf(stderr, "[ perf record: Woken up %ld times to write data ]\n", waking);
 
 	/*
@@ -820,6 +823,7 @@ static const struct option options[] = {
 		    "do call-graph (stack chain/backtrace) recording"),
 	OPT_INCR('v', "verbose", &verbose,
 		    "be more verbose (show counter open errors, etc)"),
+	OPT_BOOLEAN('q', "quiet", &quiet, "don't print any message"),
 	OPT_BOOLEAN('s', "stat", &inherit_stat,
 		    "per thread counts"),
 	OPT_BOOLEAN('d', "data", &sample_address,

tools/perf/builtin-trace.c

@@ -46,9 +46,6 @@ static struct scripting_ops	*scripting_ops;
 
 static void setup_scripting(void)
 {
-	/* make sure PERF_EXEC_PATH is set for scripts */
-	perf_set_argv_exec_path(perf_exec_path());
-
 	setup_perl_scripting();
 	setup_python_scripting();
 
@@ -285,7 +282,7 @@ static int parse_scriptname(const struct option *opt __used,
 		script++;
 	} else {
 		script = str;
-		ext = strchr(script, '.');
+		ext = strrchr(script, '.');
 		if (!ext) {
 			fprintf(stderr, "invalid script extension");
 			return -1;
@@ -593,6 +590,9 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used)
 		suffix = REPORT_SUFFIX;
 	}
 
+	/* make sure PERF_EXEC_PATH is set for scripts */
+	perf_set_argv_exec_path(perf_exec_path());
+
 	if (!suffix && argc >= 2 && strncmp(argv[1], "-", strlen("-")) != 0) {
 		char *record_script_path, *report_script_path;
 		int live_pipe[2];
@@ -625,12 +625,13 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used)
 			dup2(live_pipe[1], 1);
 			close(live_pipe[0]);
 
-			__argv = malloc(5 * sizeof(const char *));
+			__argv = malloc(6 * sizeof(const char *));
 			__argv[0] = "/bin/sh";
 			__argv[1] = record_script_path;
-			__argv[2] = "-o";
-			__argv[3] = "-";
-			__argv[4] = NULL;
+			__argv[2] = "-q";
+			__argv[3] = "-o";
+			__argv[4] = "-";
+			__argv[5] = NULL;
 
 			execvp("/bin/sh", (char **)__argv);
 			exit(-1);

tools/perf/scripts/perl/bin/failed-syscalls-report

@@ -7,4 +7,4 @@ if [ $# -gt 0 ] ; then
 	shift
     fi
 fi
-perf trace $@ -s ~/libexec/perf-core/scripts/perl/failed-syscalls.pl $comm
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/failed-syscalls.pl $comm

tools/perf/scripts/perl/bin/rw-by-file-report

@@ -7,7 +7,7 @@ if [ $# -lt 1 ] ; then
 fi
 comm=$1
 shift
-perf trace $@ -s ~/libexec/perf-core/scripts/perl/rw-by-file.pl $comm
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/rw-by-file.pl $comm
 
 
 

tools/perf/scripts/perl/bin/rw-by-pid-report

 #!/bin/bash
 # description: system-wide r/w activity
-perf trace $@ -s ~/libexec/perf-core/scripts/perl/rw-by-pid.pl
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/rw-by-pid.pl
 
 
 

tools/perf/scripts/perl/bin/rwtop-report

@@ -17,7 +17,7 @@ if [ "$n_args" -gt 0 ] ; then
     interval=$1
     shift
 fi
-perf trace $@ -s ~/libexec/perf-core/scripts/perl/rwtop.pl $interval
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/rwtop.pl $interval
 
 
 

tools/perf/scripts/perl/bin/wakeup-latency-report

 #!/bin/bash
 # description: system-wide min/max/avg wakeup latency
-perf trace $@ -s ~/libexec/perf-core/scripts/perl/wakeup-latency.pl
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/wakeup-latency.pl
 
 
 

tools/perf/scripts/perl/bin/workqueue-stats-report

 #!/bin/bash
 # description: workqueue stats (ins/exe/create/destroy)
-perf trace $@ -s ~/libexec/perf-core/scripts/perl/workqueue-stats.pl
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/workqueue-stats.pl
 
 
 

tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py

@@ -6,6 +6,14 @@
 # Public License ("GPL") version 2 as published by the Free Software
 # Foundation.
 
+import errno, os
+
+FUTEX_WAIT = 0
+FUTEX_WAKE = 1
+FUTEX_PRIVATE_FLAG = 128
+FUTEX_CLOCK_REALTIME = 256
+FUTEX_CMD_MASK = ~(FUTEX_PRIVATE_FLAG | FUTEX_CLOCK_REALTIME)
+
 NSECS_PER_SEC    = 1000000000
 
 def avg(total, n):
@@ -24,5 +32,55 @@ def nsecs_str(nsecs):
     str = "%5u.%09u" % (nsecs_secs(nsecs), nsecs_nsecs(nsecs)),
     return str
 
+def add_stats(dict, key, value):
+	if not dict.has_key(key):
+		dict[key] = (value, value, value, 1)
+	else:
+		min, max, avg, count = dict[key]
+		if value < min:
+			min = value
+		if value > max:
+			max = value
+		avg = (avg + value) / 2
+		dict[key] = (min, max, avg, count + 1)
+
 def clear_term():
     print("\x1b[H\x1b[2J")
+
+audit_package_warned = False
+
+try:
+	import audit
+	machine_to_id = {
+		'x86_64': audit.MACH_86_64,
+		'alpha'	: audit.MACH_ALPHA,
+		'ia64'	: audit.MACH_IA64,
+		'ppc'	: audit.MACH_PPC,
+		'ppc64'	: audit.MACH_PPC64,
+		's390'	: audit.MACH_S390,
+		's390x'	: audit.MACH_S390X,
+		'i386'	: audit.MACH_X86,
+		'i586'	: audit.MACH_X86,
+		'i686'	: audit.MACH_X86,
+	}
+	try:
+		machine_to_id['armeb'] = audit.MACH_ARMEB
+	except:
+		pass
+	machine_id = machine_to_id[os.uname()[4]]
+except:
+	if not audit_package_warned:
+		audit_package_warned = True
+		print "Install the audit-libs-python package to get syscall names"
+
+def syscall_name(id):
+	try:
+		return audit.audit_syscall_to_name(id, machine_id)
+	except:
+		return str(id)
+
+def strerror(nr):
+	try:
+		return errno.errorcode[abs(nr)]
+	except:
+		return "Unknown %d errno" % nr

tools/perf/scripts/python/bin/failed-syscalls-by-pid-report

@@ -7,4 +7,4 @@ if [ $# -gt 0 ] ; then
 	shift
     fi
 fi
-perf trace $@ -s ~/libexec/perf-core/scripts/python/failed-syscalls-by-pid.py $comm
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/failed-syscalls-by-pid.py $comm

tools/perf/scripts/python/bin/futex-contention-record

+#!/bin/bash
+perf record -a -e syscalls:sys_enter_futex -e syscalls:sys_exit_futex $@

tools/perf/scripts/python/bin/futex-contention-report

+#!/bin/bash
+# description: futext contention measurement
+
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/futex-contention.py

tools/perf/scripts/python/bin/netdev-times-report

@@ -2,4 +2,4 @@
 # description: display a process of packet and processing time
 # args: [tx] [rx] [dev=] [debug]
 
-perf trace -s ~/libexec/perf-core/scripts/python/netdev-times.py $@
+perf trace -s "$PERF_EXEC_PATH"/scripts/python/netdev-times.py $@

tools/perf/scripts/python/bin/sched-migration-report

 #!/bin/bash
 # description: sched migration overview
-perf trace $@ -s ~/libexec/perf-core/scripts/python/sched-migration.py
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/sched-migration.py

tools/perf/scripts/python/bin/sctop-report

@@ -21,4 +21,4 @@ elif [ "$n_args" -gt 0 ] ; then
     interval=$1
     shift
 fi
-perf trace $@ -s ~/libexec/perf-core/scripts/python/sctop.py $comm $interval
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/sctop.py $comm $interval

tools/perf/scripts/python/bin/syscall-counts-by-pid-report

@@ -7,4 +7,4 @@ if [ $# -gt 0 ] ; then
 	shift
     fi
 fi
-perf trace $@ -s ~/libexec/perf-core/scripts/python/syscall-counts-by-pid.py $comm
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/syscall-counts-by-pid.py $comm

tools/perf/scripts/python/bin/syscall-counts-report

@@ -7,4 +7,4 @@ if [ $# -gt 0 ] ; then
 	shift
     fi
 fi
-perf trace $@ -s ~/libexec/perf-core/scripts/python/syscall-counts.py $comm
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/syscall-counts.py $comm

tools/perf/scripts/python/failed-syscalls-by-pid.py

@@ -13,21 +13,26 @@ sys.path.append(os.environ['PERF_EXEC_PATH'] + \
 
 from perf_trace_context import *
 from Core import *
+from Util import *
 
-usage = "perf trace -s syscall-counts-by-pid.py [comm]\n";
+usage = "perf trace -s syscall-counts-by-pid.py [comm|pid]\n";
 
 for_comm = None
+for_pid = None
 
 if len(sys.argv) > 2:
 	sys.exit(usage)
 
 if len(sys.argv) > 1:
+	try:
+		for_pid = int(sys.argv[1])
+	except:
 		for_comm = sys.argv[1]
 
 syscalls = autodict()
 
 def trace_begin():
-	pass
+	print "Press control+C to stop and show the summary"
 
 def trace_end():
 	print_error_totals()
@@ -35,8 +40,8 @@ def trace_end():
 def raw_syscalls__sys_exit(event_name, context, common_cpu,
 	common_secs, common_nsecs, common_pid, common_comm,
 	id, ret):
-	if for_comm is not None:
-		if common_comm != for_comm:
+	if (for_comm and common_comm != for_comm) or \
+	   (for_pid  and common_pid  != for_pid ):
 		return
 
 	if ret < 0:
@@ -62,7 +67,7 @@ def print_error_totals():
 		    print "\n%s [%d]\n" % (comm, pid),
 		    id_keys = syscalls[comm][pid].keys()
 		    for id in id_keys:
-			    print "  syscall: %-16d\n" % (id),
+			    print "  syscall: %-16s\n" % syscall_name(id),
 			    ret_keys = syscalls[comm][pid][id].keys()
 			    for ret, val in sorted(syscalls[comm][pid][id].iteritems(), key = lambda(k, v): (v, k),  reverse = True):
-				    print "    err = %-20d  %10d\n" % (ret, val),
+				    print "    err = %-20s  %10d\n" % (strerror(ret), val),

tools/perf/scripts/python/futex-contention.py

+# futex contention
+# (c) 2010, Arnaldo Carvalho de Melo <acme@redhat.com>
+# Licensed under the terms of the GNU GPL License version 2
+#
+# Translation of:
+#
+# http://sourceware.org/systemtap/wiki/WSFutexContention
+#
+# to perf python scripting.
+#
+# Measures futex contention
+
+import os, sys
+sys.path.append(os.environ['PERF_EXEC_PATH'] + '/scripts/python/Perf-Trace-Util/lib/Perf/Trace')
+from Util import *
+
+process_names = {}
+thread_thislock = {}
+thread_blocktime = {}
+
+lock_waits = {} # long-lived stats on (tid,lock) blockage elapsed time
+process_names = {} # long-lived pid-to-execname mapping
+
+def syscalls__sys_enter_futex(event, ctxt, cpu, s, ns, tid, comm,
+			      nr, uaddr, op, val, utime, uaddr2, val3):
+	cmd = op & FUTEX_CMD_MASK
+	if cmd != FUTEX_WAIT:
+		return # we don't care about originators of WAKE events
+
+	process_names[tid] = comm
+	thread_thislock[tid] = uaddr
+	thread_blocktime[tid] = nsecs(s, ns)
+
+def syscalls__sys_exit_futex(event, ctxt, cpu, s, ns, tid, comm,
+			     nr, ret):
+	if thread_blocktime.has_key(tid):
+		elapsed = nsecs(s, ns) - thread_blocktime[tid]
+		add_stats(lock_waits, (tid, thread_thislock[tid]), elapsed)
+		del thread_blocktime[tid]
+		del thread_thislock[tid]
+
+def trace_begin():
+	print "Press control+C to stop and show the summary"
+
+def trace_end():
+	for (tid, lock) in lock_waits:
+		min, max, avg, count = lock_waits[tid, lock]
+		print "%s[%d] lock %x contended %d times, %d avg ns" % \
+		      (process_names[tid], tid, lock, count, avg)
+

tools/perf/scripts/python/sctop.py

@@ -8,10 +8,7 @@
 # will be refreshed every [interval] seconds.  The default interval is
 # 3 seconds.
 
-import thread
-import time
-import os
-import sys
+import os, sys, thread, time
 
 sys.path.append(os.environ['PERF_EXEC_PATH'] + \
 	'/scripts/python/Perf-Trace-Util/lib/Perf/Trace')
@@ -20,7 +17,7 @@ from perf_trace_context import *
 from Core import *
 from Util import *
 
-usage = "perf trace -s syscall-counts.py [comm] [interval]\n";
+usage = "perf trace -s sctop.py [comm] [interval]\n";
 
 for_comm = None
 default_interval = 3
@@ -71,7 +68,7 @@ def print_syscall_totals(interval):
 		for id, val in sorted(syscalls.iteritems(), key = lambda(k, v): (v, k), \
 					      reverse = True):
 			try:
-				print "%-40d  %10d\n" % (id, val),
+				print "%-40s  %10d\n" % (syscall_name(id), val),
 			except TypeError:
 				pass
 		syscalls.clear()

tools/perf/scripts/python/syscall-counts-by-pid.py

@@ -5,29 +5,33 @@
 # Displays system-wide system call totals, broken down by syscall.
 # If a [comm] arg is specified, only syscalls called by [comm] are displayed.
 
-import os
-import sys
+import os, sys
 
 sys.path.append(os.environ['PERF_EXEC_PATH'] + \
 	'/scripts/python/Perf-Trace-Util/lib/Perf/Trace')
 
 from perf_trace_context import *
 from Core import *
+from Util import syscall_name
 
 usage = "perf trace -s syscall-counts-by-pid.py [comm]\n";
 
 for_comm = None
+for_pid = None
 
 if len(sys.argv) > 2:
 	sys.exit(usage)
 
 if len(sys.argv) > 1:
+	try:
+		for_pid = int(sys.argv[1])
+	except:
 		for_comm = sys.argv[1]
 
 syscalls = autodict()
 
 def trace_begin():
-	pass
+	print "Press control+C to stop and show the summary"
 
 def trace_end():
 	print_syscall_totals()
@@ -35,8 +39,9 @@ def trace_end():
 def raw_syscalls__sys_enter(event_name, context, common_cpu,
 	common_secs, common_nsecs, common_pid, common_comm,
 	id, args):
-	if for_comm is not None:
-		if common_comm != for_comm:
+
+	if (for_comm and common_comm != for_comm) or \
+	   (for_pid  and common_pid  != for_pid ):
 		return
 	try:
 		syscalls[common_comm][common_pid][id] += 1
@@ -61,4 +66,4 @@ def print_syscall_totals():
 		    id_keys = syscalls[comm][pid].keys()
 		    for id, val in sorted(syscalls[comm][pid].iteritems(), \
 				  key = lambda(k, v): (v, k),  reverse = True):
-			    print "  %-38d  %10d\n" % (id, val),
+			    print "  %-38s  %10d\n" % (syscall_name(id), val),

tools/perf/scripts/python/syscall-counts.py

@@ -13,6 +13,7 @@ sys.path.append(os.environ['PERF_EXEC_PATH'] + \
 
 from perf_trace_context import *
 from Core import *
+from Util import syscall_name
 
 usage = "perf trace -s syscall-counts.py [comm]\n";
 
@@ -27,7 +28,7 @@ if len(sys.argv) > 1:
 syscalls = autodict()
 
 def trace_begin():
-	pass
+	print "Press control+C to stop and show the summary"
 
 def trace_end():
 	print_syscall_totals()
@@ -55,4 +56,4 @@ def print_syscall_totals():
 
     for id, val in sorted(syscalls.iteritems(), key = lambda(k, v): (v, k), \
 				  reverse = True):
-	    print "%-40d  %10d\n" % (id, val),
+	    print "%-40s  %10d\n" % (syscall_name(id), val),

tools/perf/util/debug.c

@@ -12,8 +12,8 @@
 #include "debug.h"
 #include "util.h"
 
-int verbose = 0;
-bool dump_trace = false;
+int verbose;
+bool dump_trace = false, quiet = false;
 
 int eprintf(int level, const char *fmt, ...)
 {

tools/perf/util/debug.h

@@ -6,7 +6,7 @@
 #include "event.h"
 
 extern int verbose;
-extern bool dump_trace;
+extern bool quiet, dump_trace;
 
 int dump_printf(const char *fmt, ...) __attribute__((format(printf, 1, 2)));
 void trace_event(event_t *event);

tools/perf/util/map.h

@@ -215,6 +215,16 @@ struct symbol *map_groups__find_function_by_name(struct map_groups *self,
 	return map_groups__find_symbol_by_name(self, MAP__FUNCTION, name, mapp, filter);
 }
 
+static inline
+struct symbol *machine__find_kernel_function_by_name(struct machine *self,
+						     const char *name,
+						     struct map **mapp,
+						     symbol_filter_t filter)
+{
+	return map_groups__find_function_by_name(&self->kmaps, name, mapp,
+						 filter);
+}
+
 int map_groups__fixup_overlappings(struct map_groups *self, struct map *map,
 				   int verbose, FILE *fp);
 

tools/perf/util/probe-event.c

@@ -74,10 +74,9 @@ static int e_snprintf(char *str, size_t size, const char *format, ...)
 static char *synthesize_perf_probe_point(struct perf_probe_point *pp);
 static struct machine machine;
 
-/* Initialize symbol maps and path of vmlinux */
+/* Initialize symbol maps and path of vmlinux/modules */
 static int init_vmlinux(void)
 {
-	struct dso *kernel;
 	int ret;
 
 	symbol_conf.sort_by_name = true;
@@ -91,33 +90,61 @@ static int init_vmlinux(void)
 		goto out;
 	}
 
-	ret = machine__init(&machine, "/", 0);
+	ret = machine__init(&machine, "", HOST_KERNEL_ID);
 	if (ret < 0)
 		goto out;
 
-	kernel = dso__new_kernel(symbol_conf.vmlinux_name);
-	if (kernel == NULL)
-		die("Failed to create kernel dso.");
-
-	ret = __machine__create_kernel_maps(&machine, kernel);
-	if (ret < 0)
-		pr_debug("Failed to create kernel maps.\n");
-
+	if (machine__create_kernel_maps(&machine) < 0) {
+		pr_debug("machine__create_kernel_maps ");
+		goto out;
+	}
 out:
 	if (ret < 0)
 		pr_warning("Failed to init vmlinux path.\n");
 	return ret;
 }
 
-#ifdef DWARF_SUPPORT
-static int open_vmlinux(void)
+static struct symbol *__find_kernel_function_by_name(const char *name,
+						     struct map **mapp)
+{
+	return machine__find_kernel_function_by_name(&machine, name, mapp,
+						     NULL);
+}
+
+const char *kernel_get_module_path(const char *module)
 {
-	if (map__load(machine.vmlinux_maps[MAP__FUNCTION], NULL) < 0) {
+	struct dso *dso;
+
+	if (module) {
+		list_for_each_entry(dso, &machine.kernel_dsos, node) {
+			if (strncmp(dso->short_name + 1, module,
+				    dso->short_name_len - 2) == 0)
+				goto found;
+		}
+		pr_debug("Failed to find module %s.\n", module);
+		return NULL;
+	} else {
+		dso = machine.vmlinux_maps[MAP__FUNCTION]->dso;
+		if (dso__load_vmlinux_path(dso,
+			 machine.vmlinux_maps[MAP__FUNCTION], NULL) < 0) {
 			pr_debug("Failed to load kernel map.\n");
-		return -EINVAL;
+			return NULL;
+		}
+	}
+found:
+	return dso->long_name;
+}
+
+#ifdef DWARF_SUPPORT
+static int open_vmlinux(const char *module)
+{
+	const char *path = kernel_get_module_path(module);
+	if (!path) {
+		pr_err("Failed to find path of %s module", module ?: "kernel");
+		return -ENOENT;
 	}
-	pr_debug("Try to open %s\n", machine.vmlinux_maps[MAP__FUNCTION]->dso->long_name);
-	return open(machine.vmlinux_maps[MAP__FUNCTION]->dso->long_name, O_RDONLY);
+	pr_debug("Try to open %s\n", path);
+	return open(path, O_RDONLY);
 }
 
 /*
@@ -128,17 +155,16 @@ static int kprobe_convert_to_perf_probe(struct probe_trace_point *tp,
 					struct perf_probe_point *pp)
 {
 	struct symbol *sym;
-	int fd, ret = -ENOENT;
+	struct map *map;
+	u64 addr;
+	int ret = -ENOENT;
 
-	sym = map__find_symbol_by_name(machine.vmlinux_maps[MAP__FUNCTION],
-				       tp->symbol, NULL);
+	sym = __find_kernel_function_by_name(tp->symbol, &map);
 	if (sym) {
-		fd = open_vmlinux();
-		if (fd >= 0) {
-			ret = find_perf_probe_point(fd,
-						 sym->start + tp->offset, pp);
-			close(fd);
-		}
+		addr = map->unmap_ip(map, sym->start + tp->offset);
+		pr_debug("try to find %s+%ld@%llx\n", tp->symbol,
+			 tp->offset, addr);
+		ret = find_perf_probe_point((unsigned long)addr, pp);
 	}
 	if (ret <= 0) {
 		pr_debug("Failed to find corresponding probes from "
@@ -156,12 +182,12 @@ static int kprobe_convert_to_perf_probe(struct probe_trace_point *tp,
 /* Try to find perf_probe_event with debuginfo */
 static int try_to_find_probe_trace_events(struct perf_probe_event *pev,
 					   struct probe_trace_event **tevs,
-					   int max_tevs)
+					   int max_tevs, const char *module)
 {
 	bool need_dwarf = perf_probe_event_need_dwarf(pev);
 	int fd, ntevs;
 
-	fd = open_vmlinux();
+	fd = open_vmlinux(module);
 	if (fd < 0) {
 		if (need_dwarf) {
 			pr_warning("Failed to open debuginfo file.\n");
@@ -300,7 +326,7 @@ static int show_one_line(FILE *fp, int l, bool skip, bool show_num)
  * Show line-range always requires debuginfo to find source file and
  * line number.
  */
-int show_line_range(struct line_range *lr)
+int show_line_range(struct line_range *lr, const char *module)
 {
 	int l = 1;
 	struct line_node *ln;
@@ -313,7 +339,7 @@ int show_line_range(struct line_range *lr)
 	if (ret < 0)
 		return ret;
 
-	fd = open_vmlinux();
+	fd = open_vmlinux(module);
 	if (fd < 0) {
 		pr_warning("Failed to open debuginfo file.\n");
 		return fd;
@@ -378,11 +404,84 @@ int show_line_range(struct line_range *lr)
 	return ret;
 }
 
+static int show_available_vars_at(int fd, struct perf_probe_event *pev,
+				  int max_vls, bool externs)
+{
+	char *buf;
+	int ret, i;
+	struct str_node *node;
+	struct variable_list *vls = NULL, *vl;
+
+	buf = synthesize_perf_probe_point(&pev->point);
+	if (!buf)
+		return -EINVAL;
+	pr_debug("Searching variables at %s\n", buf);
+
+	ret = find_available_vars_at(fd, pev, &vls, max_vls, externs);
+	if (ret > 0) {
+		/* Some variables were found */
+		fprintf(stdout, "Available variables at %s\n", buf);
+		for (i = 0; i < ret; i++) {
+			vl = &vls[i];
+			/*
+			 * A probe point might be converted to
+			 * several trace points.
+			 */
+			fprintf(stdout, "\t@<%s+%lu>\n", vl->point.symbol,
+				vl->point.offset);
+			free(vl->point.symbol);
+			if (vl->vars) {
+				strlist__for_each(node, vl->vars)
+					fprintf(stdout, "\t\t%s\n", node->s);
+				strlist__delete(vl->vars);
+			} else
+				fprintf(stdout, "(No variables)\n");
+		}
+		free(vls);
+	} else
+		pr_err("Failed to find variables at %s (%d)\n", buf, ret);
+
+	free(buf);
+	return ret;
+}
+
+/* Show available variables on given probe point */
+int show_available_vars(struct perf_probe_event *pevs, int npevs,
+			int max_vls, const char *module, bool externs)
+{
+	int i, fd, ret = 0;
+
+	ret = init_vmlinux();
+	if (ret < 0)
+		return ret;
+
+	fd = open_vmlinux(module);
+	if (fd < 0) {
+		pr_warning("Failed to open debuginfo file.\n");
+		return fd;
+	}
+
+	setup_pager();
+
+	for (i = 0; i < npevs && ret >= 0; i++)
+		ret = show_available_vars_at(fd, &pevs[i], max_vls, externs);
+
+	close(fd);
+	return ret;
+}
+
 #else	/* !DWARF_SUPPORT */
 
 static int kprobe_convert_to_perf_probe(struct probe_trace_point *tp,
 					struct perf_probe_point *pp)
 {
+	struct symbol *sym;
+
+	sym = __find_kernel_function_by_name(tp->symbol, NULL);
+	if (!sym) {
+		pr_err("Failed to find symbol %s in kernel.\n", tp->symbol);
+		return -ENOENT;
+	}
 	pp->function = strdup(tp->symbol);
 	if (pp->function == NULL)
 		return -ENOMEM;
@@ -394,7 +493,7 @@ static int kprobe_convert_to_perf_probe(struct probe_trace_point *tp,
 
 static int try_to_find_probe_trace_events(struct perf_probe_event *pev,
 				struct probe_trace_event **tevs __unused,
-				int max_tevs __unused)
+				int max_tevs __unused, const char *mod __unused)
 {
 	if (perf_probe_event_need_dwarf(pev)) {
 		pr_warning("Debuginfo-analysis is not supported.\n");
@@ -403,12 +502,19 @@ static int try_to_find_probe_trace_events(struct perf_probe_event *pev,
 	return 0;
 }
 
-int show_line_range(struct line_range *lr __unused)
+int show_line_range(struct line_range *lr __unused, const char *module __unused)
 {
 	pr_warning("Debuginfo-analysis is not supported.\n");
 	return -ENOSYS;
 }
 
+int show_available_vars(struct perf_probe_event *pevs __unused,
+			int npevs __unused, int max_vls __unused,
+			const char *module __unused, bool externs __unused)
+{
+	pr_warning("Debuginfo-analysis is not supported.\n");
+	return -ENOSYS;
+}
 #endif
 
 int parse_line_range_desc(const char *arg, struct line_range *lr)
@@ -1516,14 +1622,14 @@ static int __add_probe_trace_events(struct perf_probe_event *pev,
 
 static int convert_to_probe_trace_events(struct perf_probe_event *pev,
 					  struct probe_trace_event **tevs,
-					  int max_tevs)
+					  int max_tevs, const char *module)
 {
 	struct symbol *sym;
 	int ret = 0, i;
 	struct probe_trace_event *tev;
 
 	/* Convert perf_probe_event with debuginfo */
-	ret = try_to_find_probe_trace_events(pev, tevs, max_tevs);
+	ret = try_to_find_probe_trace_events(pev, tevs, max_tevs, module);
 	if (ret != 0)
 		return ret;
 
@@ -1572,8 +1678,7 @@ static int convert_to_probe_trace_events(struct perf_probe_event *pev,
 	}
 
 	/* Currently just checking function name from symbol map */
-	sym = map__find_symbol_by_name(machine.vmlinux_maps[MAP__FUNCTION],
-				       tev->point.symbol, NULL);
+	sym = __find_kernel_function_by_name(tev->point.symbol, NULL);
 	if (!sym) {
 		pr_warning("Kernel symbol \'%s\' not found.\n",
 			   tev->point.symbol);
@@ -1596,7 +1701,7 @@ struct __event_package {
 };
 
 int add_perf_probe_events(struct perf_probe_event *pevs, int npevs,
-			  bool force_add, int max_tevs)
+			  int max_tevs, const char *module, bool force_add)
 {
 	int i, j, ret;
 	struct __event_package *pkgs;
@@ -1617,7 +1722,9 @@ int add_perf_probe_events(struct perf_probe_event *pevs, int npevs,
 		pkgs[i].pev = &pevs[i];
 		/* Convert with or without debuginfo */
 		ret  = convert_to_probe_trace_events(pkgs[i].pev,
-						      &pkgs[i].tevs, max_tevs);
+						     &pkgs[i].tevs,
+						     max_tevs,
+						     module);
 		if (ret < 0)
 			goto end;
 		pkgs[i].ntevs = ret;

tools/perf/util/probe-event.h

@@ -90,6 +90,12 @@ struct line_range {
 	struct list_head	line_list;	/* Visible lines */
 };
 
+/* List of variables */
+struct variable_list {
+	struct probe_trace_point	point;	/* Actual probepoint */
+	struct strlist			*vars;	/* Available variables */
+};
+
 /* Command string to events */
 extern int parse_perf_probe_command(const char *cmd,
 				    struct perf_probe_event *pev);
@@ -109,12 +115,18 @@ extern void clear_perf_probe_event(struct perf_probe_event *pev);
 /* Command string to line-range */
 extern int parse_line_range_desc(const char *cmd, struct line_range *lr);
 
+/* Internal use: Return kernel/module path */
+extern const char *kernel_get_module_path(const char *module);
 
 extern int add_perf_probe_events(struct perf_probe_event *pevs, int npevs,
-				 bool force_add, int max_probe_points);
+				 int max_probe_points, const char *module,
+				 bool force_add);
 extern int del_perf_probe_events(struct strlist *dellist);
 extern int show_perf_probe_events(void);
-extern int show_line_range(struct line_range *lr);
+extern int show_line_range(struct line_range *lr, const char *module);
+extern int show_available_vars(struct perf_probe_event *pevs, int npevs,
+			       int max_probe_points, const char *module,
+			       bool externs);
 
 
 /* Maximum index number of event-name postfix */

tools/perf/util/probe-finder.c

@@ -116,6 +116,101 @@ static void line_list__free(struct list_head *head)
 	}
 }
 
+/* Dwarf FL wrappers */
+
+static int __linux_kernel_find_elf(Dwfl_Module *mod,
+				   void **userdata,
+				   const char *module_name,
+				   Dwarf_Addr base,
+				   char **file_name, Elf **elfp)
+{
+	int fd;
+	const char *path = kernel_get_module_path(module_name);
+
+	if (path) {
+		fd = open(path, O_RDONLY);
+		if (fd >= 0) {
+			*file_name = strdup(path);
+			return fd;
+		}
+	}
+	/* If failed, try to call standard method */
+	return dwfl_linux_kernel_find_elf(mod, userdata, module_name, base,
+					  file_name, elfp);
+}
+
+static char *debuginfo_path;	/* Currently dummy */
+
+static const Dwfl_Callbacks offline_callbacks = {
+	.find_debuginfo = dwfl_standard_find_debuginfo,
+	.debuginfo_path = &debuginfo_path,
+
+	.section_address = dwfl_offline_section_address,
+
+	/* We use this table for core files too.  */
+	.find_elf = dwfl_build_id_find_elf,
+};
+
+static const Dwfl_Callbacks kernel_callbacks = {
+	.find_debuginfo = dwfl_standard_find_debuginfo,
+	.debuginfo_path = &debuginfo_path,
+
+	.find_elf = __linux_kernel_find_elf,
+	.section_address = dwfl_linux_kernel_module_section_address,
+};
+
+/* Get a Dwarf from offline image */
+static Dwarf *dwfl_init_offline_dwarf(int fd, Dwfl **dwflp, Dwarf_Addr *bias)
+{
+	Dwfl_Module *mod;
+	Dwarf *dbg = NULL;
+
+	if (!dwflp)
+		return NULL;
+
+	*dwflp = dwfl_begin(&offline_callbacks);
+	if (!*dwflp)
+		return NULL;
+
+	mod = dwfl_report_offline(*dwflp, "", "", fd);
+	if (!mod)
+		goto error;
+
+	dbg = dwfl_module_getdwarf(mod, bias);
+	if (!dbg) {
+error:
+		dwfl_end(*dwflp);
+		*dwflp = NULL;
+	}
+	return dbg;
+}
+
+/* Get a Dwarf from live kernel image */
+static Dwarf *dwfl_init_live_kernel_dwarf(Dwarf_Addr addr, Dwfl **dwflp,
+					  Dwarf_Addr *bias)
+{
+	Dwarf *dbg;
+
+	if (!dwflp)
+		return NULL;
+
+	*dwflp = dwfl_begin(&kernel_callbacks);
+	if (!*dwflp)
+		return NULL;
+
+	/* Load the kernel dwarves: Don't care the result here */
+	dwfl_linux_kernel_report_kernel(*dwflp);
+	dwfl_linux_kernel_report_modules(*dwflp);
+
+	dbg = dwfl_addrdwarf(*dwflp, addr, bias);
+	/* Here, check whether we could get a real dwarf */
+	if (!dbg) {
+		dwfl_end(*dwflp);
+		*dwflp = NULL;
+	}
+	return dbg;
+}
+
 /* Dwarf wrappers */
 
 /* Find the realpath of the target file. */
@@ -160,26 +255,44 @@ static bool die_compare_name(Dwarf_Die *dw_die, const char *tname)
 	return name ? (strcmp(tname, name) == 0) : false;
 }
 
-/* Get type die, but skip qualifiers and typedef */
-static Dwarf_Die *die_get_real_type(Dwarf_Die *vr_die, Dwarf_Die *die_mem)
+/* Get type die */
+static Dwarf_Die *die_get_type(Dwarf_Die *vr_die, Dwarf_Die *die_mem)
 {
 	Dwarf_Attribute attr;
-	int tag;
 
-	do {
-		if (dwarf_attr(vr_die, DW_AT_type, &attr) == NULL ||
-		    dwarf_formref_die(&attr, die_mem) == NULL)
+	if (dwarf_attr_integrate(vr_die, DW_AT_type, &attr) &&
+	    dwarf_formref_die(&attr, die_mem))
+		return die_mem;
+	else
 		return NULL;
+}
 
-		tag = dwarf_tag(die_mem);
-		vr_die = die_mem;
+/* Get a type die, but skip qualifiers */
+static Dwarf_Die *__die_get_real_type(Dwarf_Die *vr_die, Dwarf_Die *die_mem)
+{
+	int tag;
+
+	do {
+		vr_die = die_get_type(vr_die, die_mem);
+		if (!vr_die)
+			break;
+		tag = dwarf_tag(vr_die);
 	} while (tag == DW_TAG_const_type ||
 		 tag == DW_TAG_restrict_type ||
 		 tag == DW_TAG_volatile_type ||
-		 tag == DW_TAG_shared_type ||
-		 tag == DW_TAG_typedef);
+		 tag == DW_TAG_shared_type);
 
-	return die_mem;
+	return vr_die;
+}
+
+/* Get a type die, but skip qualifiers and typedef */
+static Dwarf_Die *die_get_real_type(Dwarf_Die *vr_die, Dwarf_Die *die_mem)
+{
+	do {
+		vr_die = __die_get_real_type(vr_die, die_mem);
+	} while (vr_die && dwarf_tag(vr_die) == DW_TAG_typedef);
+
+	return vr_die;
 }
 
 static bool die_is_signed_type(Dwarf_Die *tp_die)
@@ -320,25 +433,35 @@ static Dwarf_Die *die_find_inlinefunc(Dwarf_Die *sp_die, Dwarf_Addr addr,
 	return die_find_child(sp_die, __die_find_inline_cb, &addr, die_mem);
 }
 
+struct __find_variable_param {
+	const char *name;
+	Dwarf_Addr addr;
+};
+
 static int __die_find_variable_cb(Dwarf_Die *die_mem, void *data)
 {
-	const char *name = data;
+	struct __find_variable_param *fvp = data;
 	int tag;
 
 	tag = dwarf_tag(die_mem);
 	if ((tag == DW_TAG_formal_parameter ||
 	     tag == DW_TAG_variable) &&
-	    die_compare_name(die_mem, name))
+	    die_compare_name(die_mem, fvp->name))
 		return DIE_FIND_CB_FOUND;
 
+	if (dwarf_haspc(die_mem, fvp->addr))
 		return DIE_FIND_CB_CONTINUE;
+	else
+		return DIE_FIND_CB_SIBLING;
 }
 
-/* Find a variable called 'name' */
-static Dwarf_Die *die_find_variable(Dwarf_Die *sp_die, const char *name,
-				    Dwarf_Die *die_mem)
+/* Find a variable called 'name' at given address */
+static Dwarf_Die *die_find_variable_at(Dwarf_Die *sp_die, const char *name,
+				       Dwarf_Addr addr, Dwarf_Die *die_mem)
 {
-	return die_find_child(sp_die, __die_find_variable_cb, (void *)name,
+	struct __find_variable_param fvp = { .name = name, .addr = addr};
+
+	return die_find_child(sp_die, __die_find_variable_cb, (void *)&fvp,
 			      die_mem);
 }
 
@@ -361,6 +484,60 @@ static Dwarf_Die *die_find_member(Dwarf_Die *st_die, const char *name,
 			      die_mem);
 }
 
+/* Get the name of given variable DIE */
+static int die_get_typename(Dwarf_Die *vr_die, char *buf, int len)
+{
+	Dwarf_Die type;
+	int tag, ret, ret2;
+	const char *tmp = "";
+
+	if (__die_get_real_type(vr_die, &type) == NULL)
+		return -ENOENT;
+
+	tag = dwarf_tag(&type);
+	if (tag == DW_TAG_array_type || tag == DW_TAG_pointer_type)
+		tmp = "*";
+	else if (tag == DW_TAG_subroutine_type) {
+		/* Function pointer */
+		ret = snprintf(buf, len, "(function_type)");
+		return (ret >= len) ? -E2BIG : ret;
+	} else {
+		if (!dwarf_diename(&type))
+			return -ENOENT;
+		if (tag == DW_TAG_union_type)
+			tmp = "union ";
+		else if (tag == DW_TAG_structure_type)
+			tmp = "struct ";
+		/* Write a base name */
+		ret = snprintf(buf, len, "%s%s", tmp, dwarf_diename(&type));
+		return (ret >= len) ? -E2BIG : ret;
+	}
+	ret = die_get_typename(&type, buf, len);
+	if (ret > 0) {
+		ret2 = snprintf(buf + ret, len - ret, "%s", tmp);
+		ret = (ret2 >= len - ret) ? -E2BIG : ret2 + ret;
+	}
+	return ret;
+}
+
+/* Get the name and type of given variable DIE, stored as "type\tname" */
+static int die_get_varname(Dwarf_Die *vr_die, char *buf, int len)
+{
+	int ret, ret2;
+
+	ret = die_get_typename(vr_die, buf, len);
+	if (ret < 0) {
+		pr_debug("Failed to get type, make it unknown.\n");
+		ret = snprintf(buf, len, "(unknown_type)");
+	}
+	if (ret > 0) {
+		ret2 = snprintf(buf + ret, len - ret, "\t%s",
+				dwarf_diename(vr_die));
+		ret = (ret2 >= len - ret) ? -E2BIG : ret2 + ret;
+	}
+	return ret;
+}
+
 /*
  * Probe finder related functions
  */
@@ -374,8 +551,13 @@ static struct probe_trace_arg_ref *alloc_trace_arg_ref(long offs)
 	return ref;
 }
 
-/* Show a location */
-static int convert_variable_location(Dwarf_Die *vr_die, struct probe_finder *pf)
+/*
+ * Convert a location into trace_arg.
+ * If tvar == NULL, this just checks variable can be converted.
+ */
+static int convert_variable_location(Dwarf_Die *vr_die, Dwarf_Addr addr,
+				     Dwarf_Op *fb_ops,
+				     struct probe_trace_arg *tvar)
 {
 	Dwarf_Attribute attr;
 	Dwarf_Op *op;
@@ -384,20 +566,23 @@ static int convert_variable_location(Dwarf_Die *vr_die, struct probe_finder *pf)
 	Dwarf_Word offs = 0;
 	bool ref = false;
 	const char *regs;
-	struct probe_trace_arg *tvar = pf->tvar;
 	int ret;
 
+	if (dwarf_attr(vr_die, DW_AT_external, &attr) != NULL)
+		goto static_var;
+
 	/* TODO: handle more than 1 exprs */
 	if (dwarf_attr(vr_die, DW_AT_location, &attr) == NULL ||
-	    dwarf_getlocation_addr(&attr, pf->addr, &op, &nops, 1) <= 0 ||
+	    dwarf_getlocation_addr(&attr, addr, &op, &nops, 1) <= 0 ||
 	    nops == 0) {
 		/* TODO: Support const_value */
-		pr_err("Failed to find the location of %s at this address.\n"
-		       " Perhaps, it has been optimized out.\n", pf->pvar->var);
 		return -ENOENT;
 	}
 
 	if (op->atom == DW_OP_addr) {
+static_var:
+		if (!tvar)
+			return 0;
 		/* Static variables on memory (not stack), make @varname */
 		ret = strlen(dwarf_diename(vr_die));
 		tvar->value = zalloc(ret + 2);
@@ -412,14 +597,11 @@ static int convert_variable_location(Dwarf_Die *vr_die, struct probe_finder *pf)
 
 	/* If this is based on frame buffer, set the offset */
 	if (op->atom == DW_OP_fbreg) {
-		if (pf->fb_ops == NULL) {
-			pr_warning("The attribute of frame base is not "
-				   "supported.\n");
+		if (fb_ops == NULL)
 			return -ENOTSUP;
-		}
 		ref = true;
 		offs = op->number;
-		op = &pf->fb_ops[0];
+		op = &fb_ops[0];
 	}
 
 	if (op->atom >= DW_OP_breg0 && op->atom <= DW_OP_breg31) {
@@ -435,13 +617,18 @@ static int convert_variable_location(Dwarf_Die *vr_die, struct probe_finder *pf)
 	} else if (op->atom == DW_OP_regx) {
 		regn = op->number;
 	} else {
-		pr_warning("DW_OP %x is not supported.\n", op->atom);
+		pr_debug("DW_OP %x is not supported.\n", op->atom);
 		return -ENOTSUP;
 	}
 
+	if (!tvar)
+		return 0;
+
 	regs = get_arch_regstr(regn);
 	if (!regs) {
-		pr_warning("Mapping for DWARF register number %u missing on this architecture.", regn);
+		/* This should be a bug in DWARF or this tool */
+		pr_warning("Mapping for DWARF register number %u "
+			   "missing on this architecture.", regn);
 		return -ERANGE;
 	}
 
@@ -666,8 +853,14 @@ static int convert_variable(Dwarf_Die *vr_die, struct probe_finder *pf)
 	pr_debug("Converting variable %s into trace event.\n",
 		 dwarf_diename(vr_die));
 
-	ret = convert_variable_location(vr_die, pf);
-	if (ret == 0 && pf->pvar->field) {
+	ret = convert_variable_location(vr_die, pf->addr, pf->fb_ops,
+					pf->tvar);
+	if (ret == -ENOENT)
+		pr_err("Failed to find the location of %s at this address.\n"
+		       " Perhaps, it has been optimized out.\n", pf->pvar->var);
+	else if (ret == -ENOTSUP)
+		pr_err("Sorry, we don't support this variable location yet.\n");
+	else if (pf->pvar->field) {
 		ret = convert_variable_fields(vr_die, pf->pvar->var,
 					      pf->pvar->field, &pf->tvar->ref,
 					      &die_mem);
@@ -722,56 +915,39 @@ static int find_variable(Dwarf_Die *sp_die, struct probe_finder *pf)
 	pr_debug("Searching '%s' variable in context.\n",
 		 pf->pvar->var);
 	/* Search child die for local variables and parameters. */
-	if (die_find_variable(sp_die, pf->pvar->var, &vr_die))
+	if (die_find_variable_at(sp_die, pf->pvar->var, pf->addr, &vr_die))
 		ret = convert_variable(&vr_die, pf);
 	else {
 		/* Search upper class */
 		nscopes = dwarf_getscopes_die(sp_die, &scopes);
-		if (nscopes > 0) {
-			ret = dwarf_getscopevar(scopes, nscopes, pf->pvar->var,
-						0, NULL, 0, 0, &vr_die);
-			if (ret >= 0)
+		while (nscopes-- > 1) {
+			pr_debug("Searching variables in %s\n",
+				 dwarf_diename(&scopes[nscopes]));
+			/* We should check this scope, so give dummy address */
+			if (die_find_variable_at(&scopes[nscopes],
+						 pf->pvar->var, 0,
+						 &vr_die)) {
 				ret = convert_variable(&vr_die, pf);
-			else
-				ret = -ENOENT;
+				goto found;
+			}
+		}
+		if (scopes)
 			free(scopes);
-		} else
 		ret = -ENOENT;
 	}
+found:
 	if (ret < 0)
 		pr_warning("Failed to find '%s' in this function.\n",
 			   pf->pvar->var);
 	return ret;
 }
 
-/* Show a probe point to output buffer */
-static int convert_probe_point(Dwarf_Die *sp_die, struct probe_finder *pf)
+/* Convert subprogram DIE to trace point */
+static int convert_to_trace_point(Dwarf_Die *sp_die, Dwarf_Addr paddr,
+				  bool retprobe, struct probe_trace_point *tp)
 {
-	struct probe_trace_event *tev;
 	Dwarf_Addr eaddr;
-	Dwarf_Die die_mem;
 	const char *name;
-	int ret, i;
-	Dwarf_Attribute fb_attr;
-	size_t nops;
-
-	if (pf->ntevs == pf->max_tevs) {
-		pr_warning("Too many( > %d) probe point found.\n",
-			   pf->max_tevs);
-		return -ERANGE;
-	}
-	tev = &pf->tevs[pf->ntevs++];
-
-	/* If no real subprogram, find a real one */
-	if (!sp_die || dwarf_tag(sp_die) != DW_TAG_subprogram) {
-		sp_die = die_find_real_subprogram(&pf->cu_die,
-						 pf->addr, &die_mem);
-		if (!sp_die) {
-			pr_warning("Failed to find probe point in any "
-				   "functions.\n");
-			return -ENOENT;
-		}
-	}
 
 	/* Copy the name of probe point */
 	name = dwarf_diename(sp_die);
@@ -781,26 +957,45 @@ static int convert_probe_point(Dwarf_Die *sp_die, struct probe_finder *pf)
 				   dwarf_diename(sp_die));
 			return -ENOENT;
 		}
-		tev->point.symbol = strdup(name);
-		if (tev->point.symbol == NULL)
+		tp->symbol = strdup(name);
+		if (tp->symbol == NULL)
 			return -ENOMEM;
-		tev->point.offset = (unsigned long)(pf->addr - eaddr);
+		tp->offset = (unsigned long)(paddr - eaddr);
 	} else
 		/* This function has no name. */
-		tev->point.offset = (unsigned long)pf->addr;
+		tp->offset = (unsigned long)paddr;
 
 	/* Return probe must be on the head of a subprogram */
-	if (pf->pev->point.retprobe) {
-		if (tev->point.offset != 0) {
+	if (retprobe) {
+		if (eaddr != paddr) {
 			pr_warning("Return probe must be on the head of"
 				   " a real function\n");
 			return -EINVAL;
 		}
-		tev->point.retprobe = true;
+		tp->retprobe = true;
 	}
 
-	pr_debug("Probe point found: %s+%lu\n", tev->point.symbol,
-		 tev->point.offset);
+	return 0;
+}
+
+/* Call probe_finder callback with real subprogram DIE */
+static int call_probe_finder(Dwarf_Die *sp_die, struct probe_finder *pf)
+{
+	Dwarf_Die die_mem;
+	Dwarf_Attribute fb_attr;
+	size_t nops;
+	int ret;
+
+	/* If no real subprogram, find a real one */
+	if (!sp_die || dwarf_tag(sp_die) != DW_TAG_subprogram) {
+		sp_die = die_find_real_subprogram(&pf->cu_die,
+						  pf->addr, &die_mem);
+		if (!sp_die) {
+			pr_warning("Failed to find probe point in any "
+				   "functions.\n");
+			return -ENOENT;
+		}
+	}
 
 	/* Get the frame base attribute/ops */
 	dwarf_attr(sp_die, DW_AT_frame_base, &fb_attr);
@@ -820,22 +1015,13 @@ static int convert_probe_point(Dwarf_Die *sp_die, struct probe_finder *pf)
 #endif
 	}
 
-	/* Find each argument */
-	tev->nargs = pf->pev->nargs;
-	tev->args = zalloc(sizeof(struct probe_trace_arg) * tev->nargs);
-	if (tev->args == NULL)
-		return -ENOMEM;
-	for (i = 0; i < pf->pev->nargs; i++) {
-		pf->pvar = &pf->pev->args[i];
-		pf->tvar = &tev->args[i];
-		ret = find_variable(sp_die, pf);
-		if (ret != 0)
-			return ret;
-	}
+	/* Call finder's callback handler */
+	ret = pf->callback(sp_die, pf);
 
 	/* *pf->fb_ops will be cached in libdw. Don't free it. */
 	pf->fb_ops = NULL;
-	return 0;
+
+	return ret;
 }
 
 /* Find probe point from its line number */
@@ -871,7 +1057,7 @@ static int find_probe_point_by_line(struct probe_finder *pf)
 			 (int)i, lineno, (uintmax_t)addr);
 		pf->addr = addr;
 
-		ret = convert_probe_point(NULL, pf);
+		ret = call_probe_finder(NULL, pf);
 		/* Continuing, because target line might be inlined. */
 	}
 	return ret;
@@ -984,7 +1170,7 @@ static int find_probe_point_lazy(Dwarf_Die *sp_die, struct probe_finder *pf)
 			 (int)i, lineno, (unsigned long long)addr);
 		pf->addr = addr;
 
-		ret = convert_probe_point(sp_die, pf);
+		ret = call_probe_finder(sp_die, pf);
 		/* Continuing, because target line might be inlined. */
 	}
 	/* TODO: deallocate lines, but how? */
@@ -1019,7 +1205,7 @@ static int probe_point_inline_cb(Dwarf_Die *in_die, void *data)
 		pr_debug("found inline addr: 0x%jx\n",
 			 (uintmax_t)pf->addr);
 
-		param->retval = convert_probe_point(in_die, pf);
+		param->retval = call_probe_finder(in_die, pf);
 		if (param->retval < 0)
 			return DWARF_CB_ABORT;
 	}
@@ -1057,7 +1243,7 @@ static int probe_point_search_cb(Dwarf_Die *sp_die, void *data)
 			}
 			pf->addr += pp->offset;
 			/* TODO: Check the address in this function */
-			param->retval = convert_probe_point(sp_die, pf);
+			param->retval = call_probe_finder(sp_die, pf);
 		}
 	} else {
 		struct dwarf_callback_param _param = {.data = (void *)pf,
@@ -1079,90 +1265,276 @@ static int find_probe_point_by_func(struct probe_finder *pf)
 	return _param.retval;
 }
 
-/* Find probe_trace_events specified by perf_probe_event from debuginfo */
-int find_probe_trace_events(int fd, struct perf_probe_event *pev,
-			     struct probe_trace_event **tevs, int max_tevs)
+/* Find probe points from debuginfo */
+static int find_probes(int fd, struct probe_finder *pf)
 {
-	struct probe_finder pf = {.pev = pev, .max_tevs = max_tevs};
-	struct perf_probe_point *pp = &pev->point;
+	struct perf_probe_point *pp = &pf->pev->point;
 	Dwarf_Off off, noff;
 	size_t cuhl;
 	Dwarf_Die *diep;
-	Dwarf *dbg;
+	Dwarf *dbg = NULL;
+	Dwfl *dwfl;
+	Dwarf_Addr bias;	/* Currently ignored */
 	int ret = 0;
 
-	pf.tevs = zalloc(sizeof(struct probe_trace_event) * max_tevs);
-	if (pf.tevs == NULL)
-		return -ENOMEM;
-	*tevs = pf.tevs;
-	pf.ntevs = 0;
-
-	dbg = dwarf_begin(fd, DWARF_C_READ);
+	dbg = dwfl_init_offline_dwarf(fd, &dwfl, &bias);
 	if (!dbg) {
 		pr_warning("No dwarf info found in the vmlinux - "
 			"please rebuild with CONFIG_DEBUG_INFO=y.\n");
-		free(pf.tevs);
-		*tevs = NULL;
 		return -EBADF;
 	}
 
 #if _ELFUTILS_PREREQ(0, 142)
 	/* Get the call frame information from this dwarf */
-	pf.cfi = dwarf_getcfi(dbg);
+	pf->cfi = dwarf_getcfi(dbg);
 #endif
 
 	off = 0;
-	line_list__init(&pf.lcache);
+	line_list__init(&pf->lcache);
 	/* Loop on CUs (Compilation Unit) */
 	while (!dwarf_nextcu(dbg, off, &noff, &cuhl, NULL, NULL, NULL) &&
 	       ret >= 0) {
 		/* Get the DIE(Debugging Information Entry) of this CU */
-		diep = dwarf_offdie(dbg, off + cuhl, &pf.cu_die);
+		diep = dwarf_offdie(dbg, off + cuhl, &pf->cu_die);
 		if (!diep)
 			continue;
 
 		/* Check if target file is included. */
 		if (pp->file)
-			pf.fname = cu_find_realpath(&pf.cu_die, pp->file);
+			pf->fname = cu_find_realpath(&pf->cu_die, pp->file);
 		else
-			pf.fname = NULL;
+			pf->fname = NULL;
 
-		if (!pp->file || pf.fname) {
+		if (!pp->file || pf->fname) {
 			if (pp->function)
-				ret = find_probe_point_by_func(&pf);
+				ret = find_probe_point_by_func(pf);
 			else if (pp->lazy_line)
-				ret = find_probe_point_lazy(NULL, &pf);
+				ret = find_probe_point_lazy(NULL, pf);
 			else {
-				pf.lno = pp->line;
-				ret = find_probe_point_by_line(&pf);
+				pf->lno = pp->line;
+				ret = find_probe_point_by_line(pf);
 			}
 		}
 		off = noff;
 	}
-	line_list__free(&pf.lcache);
-	dwarf_end(dbg);
+	line_list__free(&pf->lcache);
+	if (dwfl)
+		dwfl_end(dwfl);
 
-	return (ret < 0) ? ret : pf.ntevs;
+	return ret;
+}
+
+/* Add a found probe point into trace event list */
+static int add_probe_trace_event(Dwarf_Die *sp_die, struct probe_finder *pf)
+{
+	struct trace_event_finder *tf =
+			container_of(pf, struct trace_event_finder, pf);
+	struct probe_trace_event *tev;
+	int ret, i;
+
+	/* Check number of tevs */
+	if (tf->ntevs == tf->max_tevs) {
+		pr_warning("Too many( > %d) probe point found.\n",
+			   tf->max_tevs);
+		return -ERANGE;
+	}
+	tev = &tf->tevs[tf->ntevs++];
+
+	ret = convert_to_trace_point(sp_die, pf->addr, pf->pev->point.retprobe,
+				     &tev->point);
+	if (ret < 0)
+		return ret;
+
+	pr_debug("Probe point found: %s+%lu\n", tev->point.symbol,
+		 tev->point.offset);
+
+	/* Find each argument */
+	tev->nargs = pf->pev->nargs;
+	tev->args = zalloc(sizeof(struct probe_trace_arg) * tev->nargs);
+	if (tev->args == NULL)
+		return -ENOMEM;
+	for (i = 0; i < pf->pev->nargs; i++) {
+		pf->pvar = &pf->pev->args[i];
+		pf->tvar = &tev->args[i];
+		ret = find_variable(sp_die, pf);
+		if (ret != 0)
+			return ret;
+	}
+
+	return 0;
+}
+
+/* Find probe_trace_events specified by perf_probe_event from debuginfo */
+int find_probe_trace_events(int fd, struct perf_probe_event *pev,
+			    struct probe_trace_event **tevs, int max_tevs)
+{
+	struct trace_event_finder tf = {
+			.pf = {.pev = pev, .callback = add_probe_trace_event},
+			.max_tevs = max_tevs};
+	int ret;
+
+	/* Allocate result tevs array */
+	*tevs = zalloc(sizeof(struct probe_trace_event) * max_tevs);
+	if (*tevs == NULL)
+		return -ENOMEM;
+
+	tf.tevs = *tevs;
+	tf.ntevs = 0;
+
+	ret = find_probes(fd, &tf.pf);
+	if (ret < 0) {
+		free(*tevs);
+		*tevs = NULL;
+		return ret;
+	}
+
+	return (ret < 0) ? ret : tf.ntevs;
+}
+
+#define MAX_VAR_LEN 64
+
+/* Collect available variables in this scope */
+static int collect_variables_cb(Dwarf_Die *die_mem, void *data)
+{
+	struct available_var_finder *af = data;
+	struct variable_list *vl;
+	char buf[MAX_VAR_LEN];
+	int tag, ret;
+
+	vl = &af->vls[af->nvls - 1];
+
+	tag = dwarf_tag(die_mem);
+	if (tag == DW_TAG_formal_parameter ||
+	    tag == DW_TAG_variable) {
+		ret = convert_variable_location(die_mem, af->pf.addr,
+						af->pf.fb_ops, NULL);
+		if (ret == 0) {
+			ret = die_get_varname(die_mem, buf, MAX_VAR_LEN);
+			pr_debug2("Add new var: %s\n", buf);
+			if (ret > 0)
+				strlist__add(vl->vars, buf);
+		}
+	}
+
+	if (af->child && dwarf_haspc(die_mem, af->pf.addr))
+		return DIE_FIND_CB_CONTINUE;
+	else
+		return DIE_FIND_CB_SIBLING;
+}
+
+/* Add a found vars into available variables list */
+static int add_available_vars(Dwarf_Die *sp_die, struct probe_finder *pf)
+{
+	struct available_var_finder *af =
+			container_of(pf, struct available_var_finder, pf);
+	struct variable_list *vl;
+	Dwarf_Die die_mem, *scopes = NULL;
+	int ret, nscopes;
+
+	/* Check number of tevs */
+	if (af->nvls == af->max_vls) {
+		pr_warning("Too many( > %d) probe point found.\n", af->max_vls);
+		return -ERANGE;
+	}
+	vl = &af->vls[af->nvls++];
+
+	ret = convert_to_trace_point(sp_die, pf->addr, pf->pev->point.retprobe,
+				     &vl->point);
+	if (ret < 0)
+		return ret;
+
+	pr_debug("Probe point found: %s+%lu\n", vl->point.symbol,
+		 vl->point.offset);
+
+	/* Find local variables */
+	vl->vars = strlist__new(true, NULL);
+	if (vl->vars == NULL)
+		return -ENOMEM;
+	af->child = true;
+	die_find_child(sp_die, collect_variables_cb, (void *)af, &die_mem);
+
+	/* Find external variables */
+	if (!af->externs)
+		goto out;
+	/* Don't need to search child DIE for externs. */
+	af->child = false;
+	nscopes = dwarf_getscopes_die(sp_die, &scopes);
+	while (nscopes-- > 1)
+		die_find_child(&scopes[nscopes], collect_variables_cb,
+			       (void *)af, &die_mem);
+	if (scopes)
+		free(scopes);
+
+out:
+	if (strlist__empty(vl->vars)) {
+		strlist__delete(vl->vars);
+		vl->vars = NULL;
+	}
+
+	return ret;
+}
+
+/* Find available variables at given probe point */
+int find_available_vars_at(int fd, struct perf_probe_event *pev,
+			   struct variable_list **vls, int max_vls,
+			   bool externs)
+{
+	struct available_var_finder af = {
+			.pf = {.pev = pev, .callback = add_available_vars},
+			.max_vls = max_vls, .externs = externs};
+	int ret;
+
+	/* Allocate result vls array */
+	*vls = zalloc(sizeof(struct variable_list) * max_vls);
+	if (*vls == NULL)
+		return -ENOMEM;
+
+	af.vls = *vls;
+	af.nvls = 0;
+
+	ret = find_probes(fd, &af.pf);
+	if (ret < 0) {
+		/* Free vlist for error */
+		while (af.nvls--) {
+			if (af.vls[af.nvls].point.symbol)
+				free(af.vls[af.nvls].point.symbol);
+			if (af.vls[af.nvls].vars)
+				strlist__delete(af.vls[af.nvls].vars);
+		}
+		free(af.vls);
+		*vls = NULL;
+		return ret;
+	}
+
+	return (ret < 0) ? ret : af.nvls;
 }
 
 /* Reverse search */
-int find_perf_probe_point(int fd, unsigned long addr,
-			  struct perf_probe_point *ppt)
+int find_perf_probe_point(unsigned long addr, struct perf_probe_point *ppt)
 {
 	Dwarf_Die cudie, spdie, indie;
-	Dwarf *dbg;
+	Dwarf *dbg = NULL;
+	Dwfl *dwfl = NULL;
 	Dwarf_Line *line;
-	Dwarf_Addr laddr, eaddr;
+	Dwarf_Addr laddr, eaddr, bias = 0;
 	const char *tmp;
 	int lineno, ret = 0;
 	bool found = false;
 
-	dbg = dwarf_begin(fd, DWARF_C_READ);
-	if (!dbg)
-		return -EBADF;
+	/* Open the live linux kernel */
+	dbg = dwfl_init_live_kernel_dwarf(addr, &dwfl, &bias);
+	if (!dbg) {
+		pr_warning("No dwarf info found in the vmlinux - "
+			"please rebuild with CONFIG_DEBUG_INFO=y.\n");
+		ret = -EINVAL;
+		goto end;
+	}
 
+	/* Adjust address with bias */
+	addr += bias;
 	/* Find cu die */
-	if (!dwarf_addrdie(dbg, (Dwarf_Addr)addr, &cudie)) {
+	if (!dwarf_addrdie(dbg, (Dwarf_Addr)addr - bias, &cudie)) {
+		pr_warning("No CU DIE is found at %lx\n", addr);
 		ret = -EINVAL;
 		goto end;
 	}
@@ -1225,7 +1597,8 @@ int find_perf_probe_point(int fd, unsigned long addr,
 	}
 
 end:
-	dwarf_end(dbg);
+	if (dwfl)
+		dwfl_end(dwfl);
 	if (ret >= 0)
 		ret = found ? 1 : 0;
 	return ret;
@@ -1358,6 +1731,9 @@ static int line_range_search_cb(Dwarf_Die *sp_die, void *data)
 	struct line_finder *lf = param->data;
 	struct line_range *lr = lf->lr;
 
+	pr_debug("find (%llx) %s\n",
+		 (unsigned long long)dwarf_dieoffset(sp_die),
+		 dwarf_diename(sp_die));
 	if (dwarf_tag(sp_die) == DW_TAG_subprogram &&
 	    die_compare_name(sp_die, lr->function)) {
 		lf->fname = dwarf_decl_file(sp_die);
@@ -1401,10 +1777,12 @@ int find_line_range(int fd, struct line_range *lr)
 	Dwarf_Off off = 0, noff;
 	size_t cuhl;
 	Dwarf_Die *diep;
-	Dwarf *dbg;
+	Dwarf *dbg = NULL;
+	Dwfl *dwfl;
+	Dwarf_Addr bias;	/* Currently ignored */
 	const char *comp_dir;
 
-	dbg = dwarf_begin(fd, DWARF_C_READ);
+	dbg = dwfl_init_offline_dwarf(fd, &dwfl, &bias);
 	if (!dbg) {
 		pr_warning("No dwarf info found in the vmlinux - "
 			"please rebuild with CONFIG_DEBUG_INFO=y.\n");
@@ -1450,8 +1828,7 @@ int find_line_range(int fd, struct line_range *lr)
 	}
 
 	pr_debug("path: %s\n", lr->path);
-	dwarf_end(dbg);
-
+	dwfl_end(dwfl);
 	return (ret < 0) ? ret : lf.found;
 }
 

tools/perf/util/probe-finder.h

@@ -22,20 +22,27 @@ extern int find_probe_trace_events(int fd, struct perf_probe_event *pev,
 				    int max_tevs);
 
 /* Find a perf_probe_point from debuginfo */
-extern int find_perf_probe_point(int fd, unsigned long addr,
+extern int find_perf_probe_point(unsigned long addr,
 				 struct perf_probe_point *ppt);
 
+/* Find a line range */
 extern int find_line_range(int fd, struct line_range *lr);
 
+/* Find available variables */
+extern int find_available_vars_at(int fd, struct perf_probe_event *pev,
+				  struct variable_list **vls, int max_points,
+				  bool externs);
+
 #include <dwarf.h>
 #include <libdw.h>
+#include <libdwfl.h>
 #include <version.h>
 
 struct probe_finder {
 	struct perf_probe_event	*pev;		/* Target probe event */
-	struct probe_trace_event *tevs;		/* Result trace events */
-	int			ntevs;		/* Number of trace events */
-	int			max_tevs;	/* Max number of trace events */
+
+	/* Callback when a probe point is found */
+	int (*callback)(Dwarf_Die *sp_die, struct probe_finder *pf);
 
 	/* For function searching */
 	int			lno;		/* Line number */
@@ -53,6 +60,22 @@ struct probe_finder {
 	struct probe_trace_arg	*tvar;		/* Current result variable */
 };
 
+struct trace_event_finder {
+	struct probe_finder	pf;
+	struct probe_trace_event *tevs;		/* Found trace events */
+	int			ntevs;		/* Number of trace events */
+	int			max_tevs;	/* Max number of trace events */
+};
+
+struct available_var_finder {
+	struct probe_finder	pf;
+	struct variable_list	*vls;		/* Found variable lists */
+	int			nvls;		/* Number of variable lists */
+	int			max_vls;	/* Max no. of variable lists */
+	bool			externs;	/* Find external vars too */
+	bool			child;		/* Search child scopes */
+};
+
 struct line_finder {
 	struct line_range	*lr;		/* Target line range */
 

tools/perf/util/ui/browser.c

-#include <slang.h>
 #include "libslang.h"
 #include <linux/compiler.h>
 #include <linux/list.h>