Merge commit 'tip/perfcounters/core' into perf-counters-for-linus

10 files changed, 556 insertions, 142 deletions

diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c
index 36c3dc7..7e346d4 100644
--- a/arch/x86/kernel/cpu/perf_counter.c
+++ b/arch/x86/kernel/cpu/perf_counter.c
@@ -66,6 +66,52 @@ static DEFINE_PER_CPU(struct cpu_hw_counters, cpu_hw_counters) = {
 };
 
 /*
+ * Not sure about some of these
+ */
+static const u64 p6_perfmon_event_map[] =
+{
+  [PERF_COUNT_HW_CPU_CYCLES]		= 0x0079,
+  [PERF_COUNT_HW_INSTRUCTIONS]		= 0x00c0,
+  [PERF_COUNT_HW_CACHE_REFERENCES]	= 0x0000,
+  [PERF_COUNT_HW_CACHE_MISSES]		= 0x0000,
+  [PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x00c4,
+  [PERF_COUNT_HW_BRANCH_MISSES]		= 0x00c5,
+  [PERF_COUNT_HW_BUS_CYCLES]		= 0x0062,
+};
+
+static u64 p6_pmu_event_map(int event)
+{
+	return p6_perfmon_event_map[event];
+}
+
+/*
+ * Counter setting that is specified not to count anything.
+ * We use this to effectively disable a counter.
+ *
+ * L2_RQSTS with 0 MESI unit mask.
+ */
+#define P6_NOP_COUNTER			0x0000002EULL
+
+static u64 p6_pmu_raw_event(u64 event)
+{
+#define P6_EVNTSEL_EVENT_MASK		0x000000FFULL
+#define P6_EVNTSEL_UNIT_MASK		0x0000FF00ULL
+#define P6_EVNTSEL_EDGE_MASK		0x00040000ULL
+#define P6_EVNTSEL_INV_MASK		0x00800000ULL
+#define P6_EVNTSEL_COUNTER_MASK		0xFF000000ULL
+
+#define P6_EVNTSEL_MASK			\
+	(P6_EVNTSEL_EVENT_MASK |	\
+	 P6_EVNTSEL_UNIT_MASK  |	\
+	 P6_EVNTSEL_EDGE_MASK  |	\
+	 P6_EVNTSEL_INV_MASK   |	\
+	 P6_EVNTSEL_COUNTER_MASK)
+
+	return event & P6_EVNTSEL_MASK;
+}
+
+
+/*
  * Intel PerfMon v3. Used on Core2 and later.
  */
 static const u64 intel_perfmon_event_map[] =
@@ -666,6 +712,7 @@ static int __hw_perf_counter_init(struct perf_counter *counter)
 {
 	struct perf_counter_attr *attr = &counter->attr;
 	struct hw_perf_counter *hwc = &counter->hw;
+	u64 config;
 	int err;
 
 	if (!x86_pmu_initialized())
@@ -718,14 +765,40 @@ static int __hw_perf_counter_init(struct perf_counter *counter)
 
 	if (attr->config >= x86_pmu.max_events)
 		return -EINVAL;
+
 	/*
 	 * The generic map:
 	 */
-	hwc->config |= x86_pmu.event_map(attr->config);
+	config = x86_pmu.event_map(attr->config);
+
+	if (config == 0)
+		return -ENOENT;
+
+	if (config == -1LL)
+		return -EINVAL;
+
+	hwc->config |= config;
 
 	return 0;
 }
 
+static void p6_pmu_disable_all(void)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	u64 val;
+
+	if (!cpuc->enabled)
+		return;
+
+	cpuc->enabled = 0;
+	barrier();
+
+	/* p6 only has one enable register */
+	rdmsrl(MSR_P6_EVNTSEL0, val);
+	val &= ~ARCH_PERFMON_EVENTSEL0_ENABLE;
+	wrmsrl(MSR_P6_EVNTSEL0, val);
+}
+
 static void intel_pmu_disable_all(void)
 {
 	wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, 0);
@@ -767,6 +840,23 @@ void hw_perf_disable(void)
 	return x86_pmu.disable_all();
 }
 
+static void p6_pmu_enable_all(void)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	unsigned long val;
+
+	if (cpuc->enabled)
+		return;
+
+	cpuc->enabled = 1;
+	barrier();
+
+	/* p6 only has one enable register */
+	rdmsrl(MSR_P6_EVNTSEL0, val);
+	val |= ARCH_PERFMON_EVENTSEL0_ENABLE;
+	wrmsrl(MSR_P6_EVNTSEL0, val);
+}
+
 static void intel_pmu_enable_all(void)
 {
 	wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl);
@@ -784,13 +874,13 @@ static void amd_pmu_enable_all(void)
 	barrier();
 
 	for (idx = 0; idx < x86_pmu.num_counters; idx++) {
+		struct perf_counter *counter = cpuc->counters[idx];
 		u64 val;
 
 		if (!test_bit(idx, cpuc->active_mask))
 			continue;
-		rdmsrl(MSR_K7_EVNTSEL0 + idx, val);
-		if (val & ARCH_PERFMON_EVENTSEL0_ENABLE)
-			continue;
+
+		val = counter->hw.config;
 		val |= ARCH_PERFMON_EVENTSEL0_ENABLE;
 		wrmsrl(MSR_K7_EVNTSEL0 + idx, val);
 	}
@@ -819,16 +909,13 @@ static inline void intel_pmu_ack_status(u64 ack)
 
 static inline void x86_pmu_enable_counter(struct hw_perf_counter *hwc, int idx)
 {
-	int err;
-	err = checking_wrmsrl(hwc->config_base + idx,
+	(void)checking_wrmsrl(hwc->config_base + idx,
 			      hwc->config | ARCH_PERFMON_EVENTSEL0_ENABLE);
 }
 
 static inline void x86_pmu_disable_counter(struct hw_perf_counter *hwc, int idx)
 {
-	int err;
-	err = checking_wrmsrl(hwc->config_base + idx,
-			      hwc->config);
+	(void)checking_wrmsrl(hwc->config_base + idx, hwc->config);
 }
 
 static inline void
@@ -836,13 +923,24 @@ intel_pmu_disable_fixed(struct hw_perf_counter *hwc, int __idx)
 {
 	int idx = __idx - X86_PMC_IDX_FIXED;
 	u64 ctrl_val, mask;
-	int err;
 
 	mask = 0xfULL << (idx * 4);
 
 	rdmsrl(hwc->config_base, ctrl_val);
 	ctrl_val &= ~mask;
-	err = checking_wrmsrl(hwc->config_base, ctrl_val);
+	(void)checking_wrmsrl(hwc->config_base, ctrl_val);
+}
+
+static inline void
+p6_pmu_disable_counter(struct hw_perf_counter *hwc, int idx)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	u64 val = P6_NOP_COUNTER;
+
+	if (cpuc->enabled)
+		val |= ARCH_PERFMON_EVENTSEL0_ENABLE;
+
+	(void)checking_wrmsrl(hwc->config_base + idx, val);
 }
 
 static inline void
@@ -943,6 +1041,19 @@ intel_pmu_enable_fixed(struct hw_perf_counter *hwc, int __idx)
 	err = checking_wrmsrl(hwc->config_base, ctrl_val);
 }
 
+static void p6_pmu_enable_counter(struct hw_perf_counter *hwc, int idx)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	u64 val;
+
+	val = hwc->config;
+	if (cpuc->enabled)
+		val |= ARCH_PERFMON_EVENTSEL0_ENABLE;
+
+	(void)checking_wrmsrl(hwc->config_base + idx, val);
+}
+
+
 static void intel_pmu_enable_counter(struct hw_perf_counter *hwc, int idx)
 {
 	if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL)) {
@@ -959,8 +1070,6 @@ static void amd_pmu_enable_counter(struct hw_perf_counter *hwc, int idx)
 
 	if (cpuc->enabled)
 		x86_pmu_enable_counter(hwc, idx);
-	else
-		x86_pmu_disable_counter(hwc, idx);
 }
 
 static int
@@ -1176,6 +1285,49 @@ static void intel_pmu_reset(void)
 	local_irq_restore(flags);
 }
 
+static int p6_pmu_handle_irq(struct pt_regs *regs)
+{
+	struct perf_sample_data data;
+	struct cpu_hw_counters *cpuc;
+	struct perf_counter *counter;
+	struct hw_perf_counter *hwc;
+	int idx, handled = 0;
+	u64 val;
+
+	data.regs = regs;
+	data.addr = 0;
+
+	cpuc = &__get_cpu_var(cpu_hw_counters);
+
+	for (idx = 0; idx < x86_pmu.num_counters; idx++) {
+		if (!test_bit(idx, cpuc->active_mask))
+			continue;
+
+		counter = cpuc->counters[idx];
+		hwc = &counter->hw;
+
+		val = x86_perf_counter_update(counter, hwc, idx);
+		if (val & (1ULL << (x86_pmu.counter_bits - 1)))
+			continue;
+
+		/*
+		 * counter overflow
+		 */
+		handled		= 1;
+		data.period	= counter->hw.last_period;
+
+		if (!x86_perf_counter_set_period(counter, hwc, idx))
+			continue;
+
+		if (perf_counter_overflow(counter, 1, &data))
+			p6_pmu_disable_counter(hwc, idx);
+	}
+
+	if (handled)
+		inc_irq_stat(apic_perf_irqs);
+
+	return handled;
+}
 
 /*
  * This handler is triggered by the local APIC, so the APIC IRQ handling
@@ -1185,14 +1337,13 @@ static int intel_pmu_handle_irq(struct pt_regs *regs)
 {
 	struct perf_sample_data data;
 	struct cpu_hw_counters *cpuc;
-	int bit, cpu, loops;
+	int bit, loops;
 	u64 ack, status;
 
 	data.regs = regs;
 	data.addr = 0;
 
-	cpu = smp_processor_id();
-	cpuc = &per_cpu(cpu_hw_counters, cpu);
+	cpuc = &__get_cpu_var(cpu_hw_counters);
 
 	perf_disable();
 	status = intel_pmu_get_status();
@@ -1249,14 +1400,13 @@ static int amd_pmu_handle_irq(struct pt_regs *regs)
 	struct cpu_hw_counters *cpuc;
 	struct perf_counter *counter;
 	struct hw_perf_counter *hwc;
-	int cpu, idx, handled = 0;
+	int idx, handled = 0;
 	u64 val;
 
 	data.regs = regs;
 	data.addr = 0;
 
-	cpu = smp_processor_id();
-	cpuc = &per_cpu(cpu_hw_counters, cpu);
+	cpuc = &__get_cpu_var(cpu_hw_counters);
 
 	for (idx = 0; idx < x86_pmu.num_counters; idx++) {
 		if (!test_bit(idx, cpuc->active_mask))
@@ -1353,6 +1503,32 @@ static __read_mostly struct notifier_block perf_counter_nmi_notifier = {
 	.priority		= 1
 };
 
+static struct x86_pmu p6_pmu = {
+	.name			= "p6",
+	.handle_irq		= p6_pmu_handle_irq,
+	.disable_all		= p6_pmu_disable_all,
+	.enable_all		= p6_pmu_enable_all,
+	.enable			= p6_pmu_enable_counter,
+	.disable		= p6_pmu_disable_counter,
+	.eventsel		= MSR_P6_EVNTSEL0,
+	.perfctr		= MSR_P6_PERFCTR0,
+	.event_map		= p6_pmu_event_map,
+	.raw_event		= p6_pmu_raw_event,
+	.max_events		= ARRAY_SIZE(p6_perfmon_event_map),
+	.max_period		= (1ULL << 31) - 1,
+	.version		= 0,
+	.num_counters		= 2,
+	/*
+	 * Counters have 40 bits implemented. However they are designed such
+	 * that bits [32-39] are sign extensions of bit 31. As such the
+	 * effective width of a counter for P6-like PMU is 32 bits only.
+	 *
+	 * See IA-32 Intel Architecture Software developer manual Vol 3B
+	 */
+	.counter_bits		= 32,
+	.counter_mask		= (1ULL << 32) - 1,
+};
+
 static struct x86_pmu intel_pmu = {
 	.name			= "Intel",
 	.handle_irq		= intel_pmu_handle_irq,
@@ -1392,6 +1568,39 @@ static struct x86_pmu amd_pmu = {
 	.max_period		= (1ULL << 47) - 1,
 };
 
+static int p6_pmu_init(void)
+{
+	int high, low;
+
+	switch (boot_cpu_data.x86_model) {
+	case 1:
+	case 3:  /* Pentium Pro */
+	case 5:
+	case 6:  /* Pentium II */
+	case 7:
+	case 8:
+	case 11: /* Pentium III */
+		break;
+	case 9:
+	case 13:
+		/* Pentium M */
+		break;
+	default:
+		pr_cont("unsupported p6 CPU model %d ",
+			boot_cpu_data.x86_model);
+		return -ENODEV;
+	}
+
+	if (!cpu_has_apic) {
+		pr_info("no Local APIC, try rebooting with lapic");
+		return -ENODEV;
+	}
+
+	x86_pmu				= p6_pmu;
+
+	return 0;
+}
+
 static int intel_pmu_init(void)
 {
 	union cpuid10_edx edx;
@@ -1400,8 +1609,14 @@ static int intel_pmu_init(void)
 	unsigned int ebx;
 	int version;
 
-	if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
+	if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) {
+		/* check for P6 processor family */
+	   if (boot_cpu_data.x86 == 6) {
+		return p6_pmu_init();
+	   } else {
 		return -ENODEV;
+	   }
+	}
 
 	/*
 	 * Check whether the Architectural PerfMon supports
diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index f7a8ab9..5c6fae4 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -146,6 +146,14 @@ static void put_ctx(struct perf_counter_context *ctx)
 	}
 }
 
+static void unclone_ctx(struct perf_counter_context *ctx)
+{
+	if (ctx->parent_ctx) {
+		put_ctx(ctx->parent_ctx);
+		ctx->parent_ctx = NULL;
+	}
+}
+
 /*
  * Get the perf_counter_context for a task and lock it.
  * This has to cope with with the fact that until it is locked,
@@ -1463,10 +1471,8 @@ static void perf_counter_enable_on_exec(struct task_struct *task)
 	/*
 	 * Unclone this context if we enabled any counter.
 	 */
-	if (enabled && ctx->parent_ctx) {
-		put_ctx(ctx->parent_ctx);
-		ctx->parent_ctx = NULL;
-	}
+	if (enabled)
+		unclone_ctx(ctx);
 
 	spin_unlock(&ctx->lock);
 
@@ -1526,7 +1532,6 @@ __perf_counter_init_context(struct perf_counter_context *ctx,
 
 static struct perf_counter_context *find_get_context(pid_t pid, int cpu)
 {
-	struct perf_counter_context *parent_ctx;
 	struct perf_counter_context *ctx;
 	struct perf_cpu_context *cpuctx;
 	struct task_struct *task;
@@ -1586,11 +1591,7 @@ static struct perf_counter_context *find_get_context(pid_t pid, int cpu)
  retry:
 	ctx = perf_lock_task_context(task, &flags);
 	if (ctx) {
-		parent_ctx = ctx->parent_ctx;
-		if (parent_ctx) {
-			put_ctx(parent_ctx);
-			ctx->parent_ctx = NULL;		/* no longer a clone */
-		}
+		unclone_ctx(ctx);
 		spin_unlock_irqrestore(&ctx->lock, flags);
 	}
 
@@ -4262,15 +4263,12 @@ void perf_counter_exit_task(struct task_struct *child)
 	 */
 	spin_lock(&child_ctx->lock);
 	child->perf_counter_ctxp = NULL;
-	if (child_ctx->parent_ctx) {
-		/*
-		 * This context is a clone; unclone it so it can't get
-		 * swapped to another process while we're removing all
-		 * the counters from it.
-		 */
-		put_ctx(child_ctx->parent_ctx);
-		child_ctx->parent_ctx = NULL;
-	}
+	/*
+	 * If this context is a clone; unclone it so it can't get
+	 * swapped to another process while we're removing all
+	 * the counters from it.
+	 */
+	unclone_ctx(child_ctx);
 	spin_unlock(&child_ctx->lock);
 	local_irq_restore(flags);
 
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 8aa3f8c..e72e931 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -24,6 +24,9 @@ OPTIONS
 --dsos=::
 	Only consider symbols in these dsos. CSV that understands
 	file://filename entries.
+-n
+--show-nr-samples
+	Show the number of samples for each symbol
 -C::
 --comms=::
 	Only consider symbols in these comms. CSV that understands
@@ -33,6 +36,18 @@ OPTIONS
 	Only consider these symbols. CSV that understands
 	file://filename entries.
 
+-w::
+--field-width=::
+	Force each column width to the provided list, for large terminal
+	readability.
+
+-t::
+--field-separator=::
+
+	Use a special separator character and don't pad with spaces, replacing
+	all occurances of this separator in symbol names (and other output)
+	with a '.' character, that thus it's the only non valid separator.
+
 SEE ALSO
 --------
 linkperf:perf-stat[1]
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 4b980cc..a118bc7 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -33,8 +33,10 @@ static char		*vmlinux = NULL;
 
 static char		default_sort_order[] = "comm,dso";
 static char		*sort_order = default_sort_order;
-static char		*dso_list_str, *comm_list_str, *sym_list_str;
+static char		*dso_list_str, *comm_list_str, *sym_list_str,
+			*col_width_list_str;
 static struct strlist	*dso_list, *comm_list, *sym_list;
+static char		*field_sep;
 
 static int		input;
 static int		show_mask = SHOW_KERNEL | SHOW_USER | SHOW_HV;
@@ -49,6 +51,7 @@ static int		verbose;
 static int		modules;
 
 static int		full_paths;
+static int		show_nr_samples;
 
 static unsigned long	page_size;
 static unsigned long	mmap_window = 32;
@@ -129,6 +132,33 @@ typedef union event_union {
 	struct read_event		read;
 } event_t;
 
+static int repsep_fprintf(FILE *fp, const char *fmt, ...)
+{
+	int n;
+	va_list ap;
+
+	va_start(ap, fmt);
+	if (!field_sep)
+		n = vfprintf(fp, fmt, ap);
+	else {
+		char *bf = NULL;
+		n = vasprintf(&bf, fmt, ap);
+		if (n > 0) {
+			char *sep = bf;
+			while (1) {
+				sep = strchr(sep, *field_sep);
+				if (sep == NULL)
+					break;
+				*sep = '.';
+			}
+		}
+		fputs(bf, fp);
+		free(bf);
+	}
+	va_end(ap);
+	return n;
+}
+
 static LIST_HEAD(dsos);
 static struct dso *kernel_dso;
 static struct dso *vdso;
@@ -360,12 +390,28 @@ static struct thread *thread__new(pid_t pid)
 	return self;
 }
 
+static unsigned int dsos__col_width,
+		    comms__col_width,
+		    threads__col_width;
+
 static int thread__set_comm(struct thread *self, const char *comm)
 {
 	if (self->comm)
 		free(self->comm);
 	self->comm = strdup(comm);
-	return self->comm ? 0 : -ENOMEM;
+	if (!self->comm)
+		return -ENOMEM;
+
+	if (!col_width_list_str && !field_sep &&
+	    (!comm_list || strlist__has_entry(comm_list, comm))) {
+		unsigned int slen = strlen(comm);
+		if (slen > comms__col_width) {
+			comms__col_width = slen;
+			threads__col_width = slen + 6;
+		}
+	}
+
+	return 0;
 }
 
 static size_t thread__fprintf(struct thread *self, FILE *fp)
@@ -536,7 +582,9 @@ struct sort_entry {
 
 	int64_t (*cmp)(struct hist_entry *, struct hist_entry *);
 	int64_t (*collapse)(struct hist_entry *, struct hist_entry *);
-	size_t	(*print)(FILE *fp, struct hist_entry *);
+	size_t	(*print)(FILE *fp, struct hist_entry *, unsigned int width);
+	unsigned int *width;
+	bool	elide;
 };
 
 static int64_t cmp_null(void *l, void *r)
@@ -558,15 +606,17 @@ sort__thread_cmp(struct hist_entry *left, struct hist_entry *right)
 }
 
 static size_t
-sort__thread_print(FILE *fp, struct hist_entry *self)
+sort__thread_print(FILE *fp, struct hist_entry *self, unsigned int width)
 {
-	return fprintf(fp, "%16s:%5d", self->thread->comm ?: "", self->thread->pid);
+	return repsep_fprintf(fp, "%*s:%5d", width - 6,
+			      self->thread->comm ?: "", self->thread->pid);
 }
 
 static struct sort_entry sort_thread = {
-	.header = "         Command:  Pid",
+	.header = "Command:  Pid",
 	.cmp	= sort__thread_cmp,
 	.print	= sort__thread_print,
+	.width	= &threads__col_width,
 };
 
 /* --sort comm */
@@ -590,16 +640,17 @@ sort__comm_collapse(struct hist_entry *left, struct hist_entry *right)
 }
 
 static size_t
-sort__comm_print(FILE *fp, struct hist_entry *self)
+sort__comm_print(FILE *fp, struct hist_entry *self, unsigned int width)
 {
-	return fprintf(fp, "%16s", self->thread->comm);
+	return repsep_fprintf(fp, "%*s", width, self->thread->comm);
 }
 
 static struct sort_entry sort_comm = {
-	.header		= "         Command",
+	.header		= "Command",
 	.cmp		= sort__comm_cmp,
 	.collapse	= sort__comm_collapse,
 	.print		= sort__comm_print,
+	.width		= &comms__col_width,
 };
 
 /* --sort dso */
@@ -617,18 +668,19 @@ sort__dso_cmp(struct hist_entry *left, struct hist_entry *right)
 }
 
 static size_t
-sort__dso_print(FILE *fp, struct hist_entry *self)
+sort__dso_print(FILE *fp, struct hist_entry *self, unsigned int width)
 {
 	if (self->dso)
-		return fprintf(fp, "%-25s", self->dso->name);
+		return repsep_fprintf(fp, "%-*s", width, self->dso->name);
 
-	return fprintf(fp, "%016llx         ", (u64)self->ip);
+	return repsep_fprintf(fp, "%*llx", width, (u64)self->ip);
 }
 
 static struct sort_entry sort_dso = {
-	.header = "Shared Object            ",
+	.header = "Shared Object",
 	.cmp	= sort__dso_cmp,
 	.print	= sort__dso_print,
+	.width	= &dsos__col_width,
 };
 
 /* --sort symbol */
@@ -648,22 +700,22 @@ sort__sym_cmp(struct hist_entry *left, struct hist_entry *right)
 }
 
 static size_t
-sort__sym_print(FILE *fp, struct hist_entry *self)
+sort__sym_print(FILE *fp, struct hist_entry *self, unsigned int width __used)
 {
 	size_t ret = 0;
 
 	if (verbose)
-		ret += fprintf(fp, "%#018llx  ", (u64)self->ip);
+		ret += repsep_fprintf(fp, "%#018llx  ", (u64)self->ip);
 
+	ret += repsep_fprintf(fp, "[%c] ", self->level);
 	if (self->sym) {
-		ret += fprintf(fp, "[%c] %s",
-			self->dso == kernel_dso ? 'k' :
-			self->dso == hypervisor_dso ? 'h' : '.', self->sym->name);
+		ret += repsep_fprintf(fp, "%s", self->sym->name);
 
 		if (self->sym->module)
-			ret += fprintf(fp, "\t[%s]", self->sym->module->name);
+			ret += repsep_fprintf(fp, "\t[%s]",
+					     self->sym->module->name);
 	} else {
-		ret += fprintf(fp, "%#016llx", (u64)self->ip);
+		ret += repsep_fprintf(fp, "%#016llx", (u64)self->ip);
 	}
 
 	return ret;
@@ -690,19 +742,19 @@ sort__parent_cmp(struct hist_entry *left, struct hist_entry *right)
 }
 
 static size_t
-sort__parent_print(FILE *fp, struct hist_entry *self)
+sort__parent_print(FILE *fp, struct hist_entry *self, unsigned int width)
 {
-	size_t ret = 0;
-
-	ret += fprintf(fp, "%-20s", self->parent ? self->parent->name : "[other]");
-
-	return ret;
+	return repsep_fprintf(fp, "%-*s", width,
+			      self->parent ? self->parent->name : "[other]");
 }
 
+static unsigned int parent_symbol__col_width;
+
 static struct sort_entry sort_parent = {
-	.header = "Parent symbol       ",
+	.header = "Parent symbol",
 	.cmp	= sort__parent_cmp,
 	.print	= sort__parent_print,
+	.width	= &parent_symbol__col_width,
 };
 
 static int sort__need_collapse = 0;
@@ -967,17 +1019,25 @@ hist_entry__fprintf(FILE *fp, struct hist_entry *self, u64 total_samples)
 		return 0;
 
 	if (total_samples)
-		ret = percent_color_fprintf(fp, "   %6.2f%%",
-				(self->count * 100.0) / total_samples);
+		ret = percent_color_fprintf(fp,
+					    field_sep ? "%.2f" : "   %6.2f%%",
+					(self->count * 100.0) / total_samples);
 	else
-		ret = fprintf(fp, "%12Ld ", self->count);
+		ret = fprintf(fp, field_sep ? "%lld" : "%12lld ", self->count);
+
+	if (show_nr_samples) {
+		if (field_sep)
+			fprintf(fp, "%c%lld", *field_sep, self->count);
+		else
+			fprintf(fp, "%11lld", self->count);
+	}
 
 	list_for_each_entry(se, &hist_entry__sort_list, list) {
-		if (exclude_other && (se == &sort_parent))
+		if (se->elide)
 			continue;
 
-		fprintf(fp, "  ");
-		ret += se->print(fp, self);
+		fprintf(fp, "%s", field_sep ?: "  ");
+		ret += se->print(fp, self, se->width ? *se->width : 0);
 	}
 
 	ret += fprintf(fp, "\n");
@@ -992,6 +1052,18 @@ hist_entry__fprintf(FILE *fp, struct hist_entry *self, u64 total_samples)
  *
  */
 
+static void dso__calc_col_width(struct dso *self)
+{
+	if (!col_width_list_str && !field_sep &&
+	    (!dso_list || strlist__has_entry(dso_list, self->name))) {
+		unsigned int slen = strlen(self->name);
+		if (slen > dsos__col_width)
+			dsos__col_width = slen;
+	}
+
+	self->slen_calculated = 1;
+}
+
 static struct symbol *
 resolve_symbol(struct thread *thread, struct map **mapp,
 	       struct dso **dsop, u64 *ipp)
@@ -1011,6 +1083,14 @@ resolve_symbol(struct thread *thread, struct map **mapp,
 
 	map = thread__find_map(thread, ip);
 	if (map != NULL) {
+		/*
+		 * We have to do this here as we may have a dso
+		 * with no symbol hit that has a name longer than
+		 * the ones with symbols sampled.
+		 */
+		if (!sort_dso.elide && !map->dso->slen_calculated)
+			dso__calc_col_width(map->dso);
+
 		if (mapp)
 			*mapp = map;
 got_map:
@@ -1282,35 +1362,67 @@ static size_t output__fprintf(FILE *fp, u64 total_samples)
 	struct sort_entry *se;
 	struct rb_node *nd;
 	size_t ret = 0;
+	unsigned int width;
+	char *col_width = col_width_list_str;
 
-	fprintf(fp, "\n");
-	fprintf(fp, "#\n");
-	fprintf(fp, "# (%Ld samples)\n", (u64)total_samples);
+	fprintf(fp, "# Samples: %Ld\n", (u64)total_samples);
 	fprintf(fp, "#\n");
 
 	fprintf(fp, "# Overhead");
+	if (show_nr_samples) {
+		if (field_sep)
+			fprintf(fp, "%cSamples", *field_sep);
+		else
+			fputs("  Samples  ", fp);
+	}
 	list_for_each_entry(se, &hist_entry__sort_list, list) {
-		if (exclude_other && (se == &sort_parent))
+		if (se->elide)
+			continue;
+		if (field_sep) {
+			fprintf(fp, "%c%s", *field_sep, se->header);
 			continue;
-		fprintf(fp, "  %s", se->header);
+		}
+		width = strlen(se->header);
+		if (se->width) {
+			if (col_width_list_str) {
+				if (col_width) {
+					*se->width = atoi(col_width);
+					col_width = strchr(col_width, ',');
+					if (col_width)
+						++col_width;
+				}
+			}
+			width = *se->width = max(*se->width, width);
+		}
+		fprintf(fp, "  %*s", width, se->header);
 	}
 	fprintf(fp, "\n");
 
+	if (field_sep)
+		goto print_entries;
+
 	fprintf(fp, "# ........");
+	if (show_nr_samples)
+		fprintf(fp, " ..........");
 	list_for_each_entry(se, &hist_entry__sort_list, list) {
 		unsigned int i;
 
-		if (exclude_other && (se == &sort_parent))
+		if (se->elide)
 			continue;
 
 		fprintf(fp, "  ");
-		for (i = 0; i < strlen(se->header); i++)
+		if (se->width)
+			width = *se->width;
+		else
+			width = strlen(se->header);
+		for (i = 0; i < width; i++)
 			fprintf(fp, ".");
 	}
 	fprintf(fp, "\n");
 
 	fprintf(fp, "#\n");
 
+print_entries:
 	for (nd = rb_first(&output_hists); nd; nd = rb_next(nd)) {
 		pos = rb_entry(nd, struct hist_entry, rb_node);
 		ret += hist_entry__fprintf(fp, pos, total_samples);
@@ -1883,6 +1995,8 @@ static const struct option options[] = {
 	OPT_STRING('k', "vmlinux", &vmlinux, "file", "vmlinux pathname"),
 	OPT_BOOLEAN('m', "modules", &modules,
 		    "load module symbols - WARNING: use only with -k and LIVE kernel"),
+	OPT_BOOLEAN('n', "show-nr-samples", &show_nr_samples,
+		    "Show a column with the number of samples"),
 	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
 		   "sort by key(s): pid, comm, dso, symbol, parent"),
 	OPT_BOOLEAN('P', "full-paths", &full_paths,
@@ -1900,6 +2014,12 @@ static const struct option options[] = {
 		   "only consider symbols in these comms"),
 	OPT_STRING('S', "symbols", &sym_list_str, "symbol[,symbol...]",
 		   "only consider these symbols"),
+	OPT_STRING('w', "column-widths", &col_width_list_str,
+		   "width[,width...]",
+		   "don't try to adjust column width, use these fixed values"),
+	OPT_STRING('t', "field-separator", &field_sep, "separator",
+		   "separator for columns, no spaces will be added between "
+		   "columns '.' is reserved."),
 	OPT_END()
 };
 
@@ -1919,7 +2039,8 @@ static void setup_sorting(void)
 }
 
 static void setup_list(struct strlist **list, const char *list_str,
-		       const char *list_name)
+		       struct sort_entry *se, const char *list_name,
+		       FILE *fp)
 {
 	if (list_str) {
 		*list = strlist__new(true, list_str);
@@ -1928,6 +2049,11 @@ static void setup_list(struct strlist **list, const char *list_str,
 				list_name);
 			exit(129);
 		}
+		if (strlist__nr_entries(*list) == 1) {
+			fprintf(fp, "# %s: %s\n", list_name,
+				strlist__entry(*list, 0)->s);
+			se->elide = true;
+		}
 	}
 }
 
@@ -1941,9 +2067,10 @@ int cmd_report(int argc, const char **argv, const char *prefix __used)
 
 	setup_sorting();
 
-	if (parent_pattern != default_parent_pattern)
+	if (parent_pattern != default_parent_pattern) {
 		sort_dimension__add("parent");
-	else
+		sort_parent.elide = 1;
+	} else
 		exclude_other = 0;
 
 	/*
@@ -1952,11 +2079,17 @@ int cmd_report(int argc, const char **argv, const char *prefix __used)
 	if (argc)
 		usage_with_options(report_usage, options);
 
-	setup_list(&dso_list, dso_list_str, "dso");
-	setup_list(&comm_list, comm_list_str, "comm");
-	setup_list(&sym_list, sym_list_str, "symbol");
-
 	setup_pager();
 
+	setup_list(&dso_list, dso_list_str, &sort_dso, "dso", stdout);
+	setup_list(&comm_list, comm_list_str, &sort_comm, "comm", stdout);
+	setup_list(&sym_list, sym_list_str, &sort_sym, "symbol", stdout);
+
+	if (field_sep && *field_sep == '.') {
+		fputs("'.' is the only non valid --field-separator argument\n",
+		      stderr);
+		exit(129);
+	}
+
 	return __cmd_report();
 }
diff --git a/tools/perf/perf.h b/tools/perf/perf.h
index 27887c9..53bb955 100644
--- a/tools/perf/perf.h
+++ b/tools/perf/perf.h
@@ -1,7 +1,13 @@
 #ifndef _PERF_PERF_H
 #define _PERF_PERF_H
 
-#if defined(__x86_64__) || defined(__i386__)
+#if defined(__i386__)
+#include "../../arch/x86/include/asm/unistd.h"
+#define rmb()		asm volatile("lock; addl $0,0(%%esp)" ::: "memory")
+#define cpu_relax()	asm volatile("rep; nop" ::: "memory");
+#endif
+
+#if defined(__x86_64__)
 #include "../../arch/x86/include/asm/unistd.h"
 #define rmb()		asm volatile("lfence" ::: "memory")
 #define cpu_relax()	asm volatile("rep; nop" ::: "memory");
diff --git a/tools/perf/util/include/linux/kernel.h b/tools/perf/util/include/linux/kernel.h
index 99c1b3d..a6b8739 100644
--- a/tools/perf/util/include/linux/kernel.h
+++ b/tools/perf/util/include/linux/kernel.h
@@ -18,4 +18,12 @@
 	(type *)((char *)__mptr - offsetof(type, member)); })
 #endif
 
+#ifndef max
+#define max(x, y) ({				\
+	typeof(x) _max1 = (x);			\
+	typeof(y) _max2 = (y);			\
+	(void) (&_max1 == &_max2);		\
+	_max1 > _max2 ? _max1 : _max2; })
+#endif
+
 #endif
diff --git a/tools/perf/util/strlist.c b/tools/perf/util/strlist.c
index 025a78e..7ad3817 100644
--- a/tools/perf/util/strlist.c
+++ b/tools/perf/util/strlist.c
@@ -64,6 +64,7 @@ int strlist__add(struct strlist *self, const char *new_entry)
 
 	rb_link_node(&sn->rb_node, parent, p);
 	rb_insert_color(&sn->rb_node, &self->entries);
+	++self->nr_entries;
 
 	return 0;
 }
@@ -155,8 +156,9 @@ struct strlist *strlist__new(bool dupstr, const char *slist)
 	struct strlist *self = malloc(sizeof(*self));
 
 	if (self != NULL) {
-		self->entries = RB_ROOT;
-		self->dupstr = dupstr;
+		self->entries	 = RB_ROOT;
+		self->dupstr	 = dupstr;
+		self->nr_entries = 0;
 		if (slist && strlist__parse_list(self, slist) != 0)
 			goto out_error;
 	}
@@ -182,3 +184,17 @@ void strlist__delete(struct strlist *self)
 		free(self);
 	}
 }
+
+struct str_node *strlist__entry(const struct strlist *self, unsigned int idx)
+{
+	struct rb_node *nd;
+
+	for (nd = rb_first(&self->entries); nd; nd = rb_next(nd)) {
+		struct str_node *pos = rb_entry(nd, struct str_node, rb_node);
+
+		if (!idx--)
+			return pos;
+	}
+
+	return NULL;
+}
diff --git a/tools/perf/util/strlist.h b/tools/perf/util/strlist.h
index 2fdcfee..921818e 100644
--- a/tools/perf/util/strlist.h
+++ b/tools/perf/util/strlist.h
@@ -11,7 +11,8 @@ struct str_node {
 
 struct strlist {
 	struct rb_root entries;
-	bool dupstr;
+	unsigned int   nr_entries;
+	bool	       dupstr;
 };
 
 struct strlist *strlist__new(bool dupstr, const char *slist);
@@ -21,11 +22,17 @@ void strlist__remove(struct strlist *self, struct str_node *sn);
 int strlist__load(struct strlist *self, const char *filename);
 int strlist__add(struct strlist *self, const char *str);
 
+struct str_node *strlist__entry(const struct strlist *self, unsigned int idx);
 bool strlist__has_entry(struct strlist *self, const char *entry);
 
 static inline bool strlist__empty(const struct strlist *self)
 {
-	return rb_first(&self->entries) == NULL;
+	return self->nr_entries == 0;
+}
+
+static inline unsigned int strlist__nr_entries(const struct strlist *self)
+{
+	return self->nr_entries;
 }
 
 int strlist__parse_list(struct strlist *self, const char *s);
diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index 4683b67..f40266b 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -65,6 +65,7 @@ struct dso *dso__new(const char *name, unsigned int sym_priv_size)
 		self->syms = RB_ROOT;
 		self->sym_priv_size = sym_priv_size;
 		self->find_symbol = dso__find_symbol;
+		self->slen_calculated = 0;
 	}
 
 	return self;
@@ -373,36 +374,61 @@ static Elf_Scn *elf_section_by_name(Elf *elf, GElf_Ehdr *ep,
 	     idx < nr_entries; \
 	     ++idx, pos = gelf_getrela(reldata, idx, &pos_mem))
 
-static int dso__synthesize_plt_symbols(struct  dso *self, Elf *elf,
-				       GElf_Ehdr *ehdr, Elf_Scn *scn_dynsym,
-				       GElf_Shdr *shdr_dynsym,
-				       size_t dynsym_idx, int verbose)
+/*
+ * We need to check if we have a .dynsym, so that we can handle the
+ * .plt, synthesizing its symbols, that aren't on the symtabs (be it
+ * .dynsym or .symtab).
+ * And always look at the original dso, not at debuginfo packages, that
+ * have the PLT data stripped out (shdr_rel_plt.sh_type == SHT_NOBITS).
+ */
+static int dso__synthesize_plt_symbols(struct  dso *self, int verbose)
 {
 	uint32_t nr_rel_entries, idx;
 	GElf_Sym sym;
 	u64 plt_offset;
 	GElf_Shdr shdr_plt;
 	struct symbol *f;
-	GElf_Shdr shdr_rel_plt;
+	GElf_Shdr shdr_rel_plt, shdr_dynsym;
 	Elf_Data *reldata, *syms, *symstrs;
-	Elf_Scn *scn_plt_rel, *scn_symstrs;
+	Elf_Scn *scn_plt_rel, *scn_symstrs, *scn_dynsym;
+	size_t dynsym_idx;
+	GElf_Ehdr ehdr;
 	char sympltname[1024];
-	int nr = 0, symidx;
+	Elf *elf;
+	int nr = 0, symidx, fd, err = 0;
+
+	fd = open(self->name, O_RDONLY);
+	if (fd < 0)
+		goto out;
+
+	elf = elf_begin(fd, ELF_C_READ_MMAP, NULL);
+	if (elf == NULL)
+		goto out_close;
+
+	if (gelf_getehdr(elf, &ehdr) == NULL)
+		goto out_elf_end;
+
+	scn_dynsym = elf_section_by_name(elf, &ehdr, &shdr_dynsym,
+					 ".dynsym", &dynsym_idx);
+	if (scn_dynsym == NULL)
+		goto out_elf_end;
 
-	scn_plt_rel = elf_section_by_name(elf, ehdr, &shdr_rel_plt,
+	scn_plt_rel = elf_section_by_name(elf, &ehdr, &shdr_rel_plt,
 					  ".rela.plt", NULL);
 	if (scn_plt_rel == NULL) {
-		scn_plt_rel = elf_section_by_name(elf, ehdr, &shdr_rel_plt,
+		scn_plt_rel = elf_section_by_name(elf, &ehdr, &shdr_rel_plt,
 						  ".rel.plt", NULL);
 		if (scn_plt_rel == NULL)
-			return 0;
+			goto out_elf_end;
 	}
 
+	err = -1;
+
 	if (shdr_rel_plt.sh_link != dynsym_idx)
-		return 0;
+		goto out_elf_end;
 
-	if (elf_section_by_name(elf, ehdr, &shdr_plt, ".plt", NULL) == NULL)
-		return 0;
+	if (elf_section_by_name(elf, &ehdr, &shdr_plt, ".plt", NULL) == NULL)
+		goto out_elf_end;
 
 	/*
 	 * Fetch the relocation section to find the indexes to the GOT
@@ -410,19 +436,19 @@ static int dso__synthesize_plt_symbols(struct  dso *self, Elf *elf,
 	 */
 	reldata = elf_getdata(scn_plt_rel, NULL);
 	if (reldata == NULL)
-		return -1;
+		goto out_elf_end;
 
 	syms = elf_getdata(scn_dynsym, NULL);
 	if (syms == NULL)
-		return -1;
+		goto out_elf_end;
 
-	scn_symstrs = elf_getscn(elf, shdr_dynsym->sh_link);
+	scn_symstrs = elf_getscn(elf, shdr_dynsym.sh_link);
 	if (scn_symstrs == NULL)
-		return -1;
+		goto out_elf_end;
 
 	symstrs = elf_getdata(scn_symstrs, NULL);
 	if (symstrs == NULL)
-		return -1;
+		goto out_elf_end;
 
 	nr_rel_entries = shdr_rel_plt.sh_size / shdr_rel_plt.sh_entsize;
 	plt_offset = shdr_plt.sh_offset;
@@ -441,7 +467,7 @@ static int dso__synthesize_plt_symbols(struct  dso *self, Elf *elf,
 			f = symbol__new(plt_offset, shdr_plt.sh_entsize,
 					sympltname, self->sym_priv_size, 0, verbose);
 			if (!f)
-				return -1;
+				goto out_elf_end;
 
 			dso__insert_symbol(self, f);
 			++nr;
@@ -459,19 +485,25 @@ static int dso__synthesize_plt_symbols(struct  dso *self, Elf *elf,
 			f = symbol__new(plt_offset, shdr_plt.sh_entsize,
 					sympltname, self->sym_priv_size, 0, verbose);
 			if (!f)
-				return -1;
+				goto out_elf_end;
 
 			dso__insert_symbol(self, f);
 			++nr;
 		}
-	} else {
-		/*
-		 * TODO: There are still one more shdr_rel_plt.sh_type
-		 * I have to investigate, but probably should be ignored.
-		 */
 	}
 
-	return nr;
+	err = 0;
+out_elf_end:
+	elf_end(elf);
+out_close:
+	close(fd);
+
+	if (err == 0)
+		return nr;
+out:
+	fprintf(stderr, "%s: problems reading %s PLT info.\n",
+		__func__, self->name);
+	return 0;
 }
 
 static int dso__load_sym(struct dso *self, int fd, const char *name,
@@ -485,9 +517,8 @@ static int dso__load_sym(struct dso *self, int fd, const char *name,
 	GElf_Shdr shdr;
 	Elf_Data *syms;
 	GElf_Sym sym;
-	Elf_Scn *sec, *sec_dynsym, *sec_strndx;
+	Elf_Scn *sec, *sec_strndx;
 	Elf *elf;
-	size_t dynsym_idx;
 	int nr = 0;
 
 	elf = elf_begin(fd, ELF_C_READ_MMAP, NULL);
@@ -504,32 +535,11 @@ static int dso__load_sym(struct dso *self, int fd, const char *name,
 		goto out_elf_end;
 	}
 
-	/*
-	 * We need to check if we have a .dynsym, so that we can handle the
-	 * .plt, synthesizing its symbols, that aren't on the symtabs (be it
-	 * .dynsym or .symtab)
-	 */
-	sec_dynsym = elf_section_by_name(elf, &ehdr, &shdr,
-					 ".dynsym", &dynsym_idx);
-	if (sec_dynsym != NULL) {
-		nr = dso__synthesize_plt_symbols(self, elf, &ehdr,
-						 sec_dynsym, &shdr,
-						 dynsym_idx, verbose);
-		if (nr < 0)
-			goto out_elf_end;
-	}
-
-	/*
-	 * But if we have a full .symtab (that is a superset of .dynsym) we
-	 * should add the symbols not in the .dynsyn
-	 */
 	sec = elf_section_by_name(elf, &ehdr, &shdr, ".symtab", NULL);
 	if (sec == NULL) {
-		if (sec_dynsym == NULL)
+		sec = elf_section_by_name(elf, &ehdr, &shdr, ".dynsym", NULL);
+		if (sec == NULL)
 			goto out_elf_end;
-
-		sec = sec_dynsym;
-		gelf_getshdr(sec, &shdr);
 	}
 
 	syms = elf_getdata(sec, NULL);
@@ -668,6 +678,11 @@ more:
 	if (!ret)
 		goto more;
 
+	if (ret > 0) {
+		int nr_plt = dso__synthesize_plt_symbols(self, verbose);
+		if (nr_plt > 0)
+			ret += nr_plt;
+	}
 out:
 	free(name);
 	return ret;
diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h
index 7918cff..2f92b21 100644
--- a/tools/perf/util/symbol.h
+++ b/tools/perf/util/symbol.h
@@ -25,6 +25,7 @@ struct dso {
 	struct symbol    *(*find_symbol)(struct dso *, u64 ip);
 	unsigned int	 sym_priv_size;
 	unsigned char	 adjust_symbols;
+	unsigned char	 slen_calculated;
 	char		 name[0];
 };