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

* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (106 commits)
  perf kvm: Fix copy & paste error in description
  perf script: Kill script_spec__delete
  perf top: Fix a memory leak
  perf stat: Introduce get_ratio_color() helper
  perf session: Remove impossible condition check
  perf tools: Fix feature-bits rework fallout, remove unused variable
  perf script: Add generic perl handler to process events
  perf tools: Use for_each_set_bit() to iterate over feature flags
  perf tools: Unify handling of features when writing feature section
  perf report: Accept fifos as input file
  perf tools: Moving code in some files
  perf tools: Fix out-of-bound access to struct perf_session
  perf tools: Continue processing header on unknown features
  perf tools: Improve macros for struct feature_ops
  perf: builtin-record: Document and check that mmap_pages must be a power of two.
  perf: builtin-record: Provide advice if mmap'ing fails with EPERM.
  perf tools: Fix truncated annotation
  perf script: look up thread using tid instead of pid
  perf tools: Look up thread names for system wide profiling
  perf tools: Fix comm for processes with named threads
  ...

21 files changed, 1406 insertions, 493 deletions

diff --git a/arch/Kconfig b/arch/Kconfig
index 4b0669cbb3b0..2505740b81d2 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -30,6 +30,10 @@ config OPROFILE_EVENT_MULTIPLEX
 config HAVE_OPROFILE
 	bool
 
+config OPROFILE_NMI_TIMER
+	def_bool y
+	depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI
+
 config KPROBES
 	bool "Kprobes"
 	depends on MODULES
diff --git a/arch/s390/oprofile/hwsampler.c b/arch/s390/oprofile/hwsampler.c
index f43c0e4282af..9daee91e6c3f 100644
--- a/arch/s390/oprofile/hwsampler.c
+++ b/arch/s390/oprofile/hwsampler.c
@@ -22,6 +22,7 @@
 #include <asm/irq.h>
 
 #include "hwsampler.h"
+#include "op_counter.h"
 
 #define MAX_NUM_SDB 511
 #define MIN_NUM_SDB 1
@@ -896,6 +897,8 @@ static void add_samples_to_oprofile(unsigned int cpu, unsigned long *sdbt,
 		if (sample_data_ptr->P == 1) {
 			/* userspace sample */
 			unsigned int pid = sample_data_ptr->prim_asn;
+			if (!counter_config.user)
+				goto skip_sample;
 			rcu_read_lock();
 			tsk = pid_task(find_vpid(pid), PIDTYPE_PID);
 			if (tsk)
@@ -903,6 +906,8 @@ static void add_samples_to_oprofile(unsigned int cpu, unsigned long *sdbt,
 			rcu_read_unlock();
 		} else {
 			/* kernelspace sample */
+			if (!counter_config.kernel)
+				goto skip_sample;
 			regs = task_pt_regs(current);
 		}
 
@@ -910,7 +915,7 @@ static void add_samples_to_oprofile(unsigned int cpu, unsigned long *sdbt,
 		oprofile_add_ext_hw_sample(sample_data_ptr->ia, regs, 0,
 				!sample_data_ptr->P, tsk);
 		mutex_unlock(&hws_sem);
-
+	skip_sample:
 		sample_data_ptr++;
 	}
 }
diff --git a/arch/s390/oprofile/init.c b/arch/s390/oprofile/init.c
index bd58b72454cf..2297be406c61 100644
--- a/arch/s390/oprofile/init.c
+++ b/arch/s390/oprofile/init.c
@@ -2,10 +2,11 @@
  * arch/s390/oprofile/init.c
  *
  * S390 Version
- *   Copyright (C) 2003 IBM Deutschland Entwicklung GmbH, IBM Corporation
+ *   Copyright (C) 2002-2011 IBM Deutschland Entwicklung GmbH, IBM Corporation
  *   Author(s): Thomas Spatzier (tspat@de.ibm.com)
  *   Author(s): Mahesh Salgaonkar (mahesh@linux.vnet.ibm.com)
  *   Author(s): Heinz Graalfs (graalfs@linux.vnet.ibm.com)
+ *   Author(s): Andreas Krebbel (krebbel@linux.vnet.ibm.com)
  *
  * @remark Copyright 2002-2011 OProfile authors
  */
@@ -14,6 +15,8 @@
 #include <linux/init.h>
 #include <linux/errno.h>
 #include <linux/fs.h>
+#include <linux/module.h>
+#include <asm/processor.h>
 
 #include "../../../drivers/oprofile/oprof.h"
 
@@ -22,6 +25,7 @@ extern void s390_backtrace(struct pt_regs * const regs, unsigned int depth);
 #ifdef CONFIG_64BIT
 
 #include "hwsampler.h"
+#include "op_counter.h"
 
 #define DEFAULT_INTERVAL	4127518
 
@@ -35,16 +39,41 @@ static unsigned long oprofile_max_interval;
 static unsigned long oprofile_sdbt_blocks = DEFAULT_SDBT_BLOCKS;
 static unsigned long oprofile_sdb_blocks = DEFAULT_SDB_BLOCKS;
 
-static int hwsampler_file;
+static int hwsampler_enabled;
 static int hwsampler_running;	/* start_mutex must be held to change */
+static int hwsampler_available;
 
 static struct oprofile_operations timer_ops;
 
+struct op_counter_config counter_config;
+
+enum __force_cpu_type {
+	reserved = 0,		/* do not force */
+	timer,
+};
+static int force_cpu_type;
+
+static int set_cpu_type(const char *str, struct kernel_param *kp)
+{
+	if (!strcmp(str, "timer")) {
+		force_cpu_type = timer;
+		printk(KERN_INFO "oprofile: forcing timer to be returned "
+		                 "as cpu type\n");
+	} else {
+		force_cpu_type = 0;
+	}
+
+	return 0;
+}
+module_param_call(cpu_type, set_cpu_type, NULL, NULL, 0);
+MODULE_PARM_DESC(cpu_type, "Force legacy basic mode sampling"
+		           "(report cpu_type \"timer\"");
+
 static int oprofile_hwsampler_start(void)
 {
 	int retval;
 
-	hwsampler_running = hwsampler_file;
+	hwsampler_running = hwsampler_enabled;
 
 	if (!hwsampler_running)
 		return timer_ops.start();
@@ -72,10 +101,16 @@ static void oprofile_hwsampler_stop(void)
 	return;
 }
 
+/*
+ * File ops used for:
+ * /dev/oprofile/0/enabled
+ * /dev/oprofile/hwsampling/hwsampler  (cpu_type = timer)
+ */
+
 static ssize_t hwsampler_read(struct file *file, char __user *buf,
 		size_t count, loff_t *offset)
 {
-	return oprofilefs_ulong_to_user(hwsampler_file, buf, count, offset);
+	return oprofilefs_ulong_to_user(hwsampler_enabled, buf, count, offset);
 }
 
 static ssize_t hwsampler_write(struct file *file, char const __user *buf,
@@ -91,6 +126,9 @@ static ssize_t hwsampler_write(struct file *file, char const __user *buf,
 	if (retval <= 0)
 		return retval;
 
+	if (val != 0 && val != 1)
+		return -EINVAL;
+
 	if (oprofile_started)
 		/*
 		 * save to do without locking as we set
@@ -99,7 +137,7 @@ static ssize_t hwsampler_write(struct file *file, char const __user *buf,
 		 */
 		return -EBUSY;
 
-	hwsampler_file = val;
+	hwsampler_enabled = val;
 
 	return count;
 }
@@ -109,38 +147,311 @@ static const struct file_operations hwsampler_fops = {
 	.write		= hwsampler_write,
 };
 
+/*
+ * File ops used for:
+ * /dev/oprofile/0/count
+ * /dev/oprofile/hwsampling/hw_interval  (cpu_type = timer)
+ *
+ * Make sure that the value is within the hardware range.
+ */
+
+static ssize_t hw_interval_read(struct file *file, char __user *buf,
+				size_t count, loff_t *offset)
+{
+	return oprofilefs_ulong_to_user(oprofile_hw_interval, buf,
+					count, offset);
+}
+
+static ssize_t hw_interval_write(struct file *file, char const __user *buf,
+				 size_t count, loff_t *offset)
+{
+	unsigned long val;
+	int retval;
+
+	if (*offset)
+		return -EINVAL;
+	retval = oprofilefs_ulong_from_user(&val, buf, count);
+	if (retval)
+		return retval;
+	if (val < oprofile_min_interval)
+		oprofile_hw_interval = oprofile_min_interval;
+	else if (val > oprofile_max_interval)
+		oprofile_hw_interval = oprofile_max_interval;
+	else
+		oprofile_hw_interval = val;
+
+	return count;
+}
+
+static const struct file_operations hw_interval_fops = {
+	.read		= hw_interval_read,
+	.write		= hw_interval_write,
+};
+
+/*
+ * File ops used for:
+ * /dev/oprofile/0/event
+ * Only a single event with number 0 is supported with this counter.
+ *
+ * /dev/oprofile/0/unit_mask
+ * This is a dummy file needed by the user space tools.
+ * No value other than 0 is accepted or returned.
+ */
+
+static ssize_t hwsampler_zero_read(struct file *file, char __user *buf,
+				    size_t count, loff_t *offset)
+{
+	return oprofilefs_ulong_to_user(0, buf, count, offset);
+}
+
+static ssize_t hwsampler_zero_write(struct file *file, char const __user *buf,
+				     size_t count, loff_t *offset)
+{
+	unsigned long val;
+	int retval;
+
+	if (*offset)
+		return -EINVAL;
+
+	retval = oprofilefs_ulong_from_user(&val, buf, count);
+	if (retval)
+		return retval;
+	if (val != 0)
+		return -EINVAL;
+	return count;
+}
+
+static const struct file_operations zero_fops = {
+	.read		= hwsampler_zero_read,
+	.write		= hwsampler_zero_write,
+};
+
+/* /dev/oprofile/0/kernel file ops.  */
+
+static ssize_t hwsampler_kernel_read(struct file *file, char __user *buf,
+				     size_t count, loff_t *offset)
+{
+	return oprofilefs_ulong_to_user(counter_config.kernel,
+					buf, count, offset);
+}
+
+static ssize_t hwsampler_kernel_write(struct file *file, char const __user *buf,
+				      size_t count, loff_t *offset)
+{
+	unsigned long val;
+	int retval;
+
+	if (*offset)
+		return -EINVAL;
+
+	retval = oprofilefs_ulong_from_user(&val, buf, count);
+	if (retval)
+		return retval;
+
+	if (val != 0 && val != 1)
+		return -EINVAL;
+
+	counter_config.kernel = val;
+
+	return count;
+}
+
+static const struct file_operations kernel_fops = {
+	.read		= hwsampler_kernel_read,
+	.write		= hwsampler_kernel_write,
+};
+
+/* /dev/oprofile/0/user file ops. */
+
+static ssize_t hwsampler_user_read(struct file *file, char __user *buf,
+				   size_t count, loff_t *offset)
+{
+	return oprofilefs_ulong_to_user(counter_config.user,
+					buf, count, offset);
+}
+
+static ssize_t hwsampler_user_write(struct file *file, char const __user *buf,
+				    size_t count, loff_t *offset)
+{
+	unsigned long val;
+	int retval;
+
+	if (*offset)
+		return -EINVAL;
+
+	retval = oprofilefs_ulong_from_user(&val, buf, count);
+	if (retval)
+		return retval;
+
+	if (val != 0 && val != 1)
+		return -EINVAL;
+
+	counter_config.user = val;
+
+	return count;
+}
+
+static const struct file_operations user_fops = {
+	.read		= hwsampler_user_read,
+	.write		= hwsampler_user_write,
+};
+
+
+/*
+ * File ops used for: /dev/oprofile/timer/enabled
+ * The value always has to be the inverted value of hwsampler_enabled. So
+ * no separate variable is created. That way we do not need locking.
+ */
+
+static ssize_t timer_enabled_read(struct file *file, char __user *buf,
+				  size_t count, loff_t *offset)
+{
+	return oprofilefs_ulong_to_user(!hwsampler_enabled, buf, count, offset);
+}
+
+static ssize_t timer_enabled_write(struct file *file, char const __user *buf,
+				   size_t count, loff_t *offset)
+{
+	unsigned long val;
+	int retval;
+
+	if (*offset)
+		return -EINVAL;
+
+	retval = oprofilefs_ulong_from_user(&val, buf, count);
+	if (retval)
+		return retval;
+
+	if (val != 0 && val != 1)
+		return -EINVAL;
+
+	/* Timer cannot be disabled without having hardware sampling.  */
+	if (val == 0 && !hwsampler_available)
+		return -EINVAL;
+
+	if (oprofile_started)
+		/*
+		 * save to do without locking as we set
+		 * hwsampler_running in start() when start_mutex is
+		 * held
+		 */
+		return -EBUSY;
+
+	hwsampler_enabled = !val;
+
+	return count;
+}
+
+static const struct file_operations timer_enabled_fops = {
+	.read		= timer_enabled_read,
+	.write		= timer_enabled_write,
+};
+
+
 static int oprofile_create_hwsampling_files(struct super_block *sb,
-						struct dentry *root)
+					    struct dentry *root)
 {
-	struct dentry *hw_dir;
+	struct dentry *dir;
+
+	dir = oprofilefs_mkdir(sb, root, "timer");
+	if (!dir)
+		return -EINVAL;
+
+	oprofilefs_create_file(sb, dir, "enabled", &timer_enabled_fops);
+
+	if (!hwsampler_available)
+		return 0;
 
 	/* reinitialize default values */
-	hwsampler_file = 1;
+	hwsampler_enabled = 1;
+	counter_config.kernel = 1;
+	counter_config.user = 1;
 
-	hw_dir = oprofilefs_mkdir(sb, root, "hwsampling");
-	if (!hw_dir)
-		return -EINVAL;
+	if (!force_cpu_type) {
+		/*
+		 * Create the counter file system.  A single virtual
+		 * counter is created which can be used to
+		 * enable/disable hardware sampling dynamically from
+		 * user space.  The user space will configure a single
+		 * counter with a single event.  The value of 'event'
+		 * and 'unit_mask' are not evaluated by the kernel code
+		 * and can only be set to 0.
+		 */
+
+		dir = oprofilefs_mkdir(sb, root, "0");
+		if (!dir)
+			return -EINVAL;
 
-	oprofilefs_create_file(sb, hw_dir, "hwsampler", &hwsampler_fops);
-	oprofilefs_create_ulong(sb, hw_dir, "hw_interval",
-				&oprofile_hw_interval);
-	oprofilefs_create_ro_ulong(sb, hw_dir, "hw_min_interval",
-				&oprofile_min_interval);
-	oprofilefs_create_ro_ulong(sb, hw_dir, "hw_max_interval",
-				&oprofile_max_interval);
-	oprofilefs_create_ulong(sb, hw_dir, "hw_sdbt_blocks",
-				&oprofile_sdbt_blocks);
+		oprofilefs_create_file(sb, dir, "enabled", &hwsampler_fops);
+		oprofilefs_create_file(sb, dir, "event", &zero_fops);
+		oprofilefs_create_file(sb, dir, "count", &hw_interval_fops);
+		oprofilefs_create_file(sb, dir, "unit_mask", &zero_fops);
+		oprofilefs_create_file(sb, dir, "kernel", &kernel_fops);
+		oprofilefs_create_file(sb, dir, "user", &user_fops);
+		oprofilefs_create_ulong(sb, dir, "hw_sdbt_blocks",
+					&oprofile_sdbt_blocks);
 
+	} else {
+		/*
+		 * Hardware sampling can be used but the cpu_type is
+		 * forced to timer in order to deal with legacy user
+		 * space tools.  The /dev/oprofile/hwsampling fs is
+		 * provided in that case.
+		 */
+		dir = oprofilefs_mkdir(sb, root, "hwsampling");
+		if (!dir)
+			return -EINVAL;
+
+		oprofilefs_create_file(sb, dir, "hwsampler",
+				       &hwsampler_fops);
+		oprofilefs_create_file(sb, dir, "hw_interval",
+				       &hw_interval_fops);
+		oprofilefs_create_ro_ulong(sb, dir, "hw_min_interval",
+					   &oprofile_min_interval);
+		oprofilefs_create_ro_ulong(sb, dir, "hw_max_interval",
+					   &oprofile_max_interval);
+		oprofilefs_create_ulong(sb, dir, "hw_sdbt_blocks",
+					&oprofile_sdbt_blocks);
+	}
 	return 0;
 }
 
 static int oprofile_hwsampler_init(struct oprofile_operations *ops)
 {
+	/*
+	 * Initialize the timer mode infrastructure as well in order
+	 * to be able to switch back dynamically.  oprofile_timer_init
+	 * is not supposed to fail.
+	 */
+	if (oprofile_timer_init(ops))
+		BUG();
+
+	memcpy(&timer_ops, ops, sizeof(timer_ops));
+	ops->create_files = oprofile_create_hwsampling_files;
+
+	/*
+	 * If the user space tools do not support newer cpu types,
+	 * the force_cpu_type module parameter
+	 * can be used to always return \"timer\" as cpu type.
+	 */
+	if (force_cpu_type != timer) {
+		struct cpuid id;
+
+		get_cpu_id (&id);
+
+		switch (id.machine) {
+		case 0x2097: case 0x2098: ops->cpu_type = "s390/z10"; break;
+		case 0x2817: case 0x2818: ops->cpu_type = "s390/z196"; break;
+		default: return -ENODEV;
+		}
+	}
+
 	if (hwsampler_setup())
 		return -ENODEV;
 
 	/*
-	 * create hwsampler files only if hwsampler_setup() succeeds.
+	 * Query the range for the sampling interval from the
+	 * hardware.
 	 */
 	oprofile_min_interval = hwsampler_query_min_interval();
 	if (oprofile_min_interval == 0)
@@ -155,23 +466,17 @@ static int oprofile_hwsampler_init(struct oprofile_operations *ops)
 	if (oprofile_hw_interval > oprofile_max_interval)
 		oprofile_hw_interval = oprofile_max_interval;
 
-	if (oprofile_timer_init(ops))
-		return -ENODEV;
-
-	printk(KERN_INFO "oprofile: using hardware sampling\n");
-
-	memcpy(&timer_ops, ops, sizeof(timer_ops));
+	printk(KERN_INFO "oprofile: System z hardware sampling "
+	       "facility found.\n");
 
 	ops->start = oprofile_hwsampler_start;
 	ops->stop = oprofile_hwsampler_stop;
-	ops->create_files = oprofile_create_hwsampling_files;
 
 	return 0;
 }
 
 static void oprofile_hwsampler_exit(void)
 {
-	oprofile_timer_exit();
 	hwsampler_shutdown();
 }
 
@@ -182,7 +487,15 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	ops->backtrace = s390_backtrace;
 
 #ifdef CONFIG_64BIT
-	return oprofile_hwsampler_init(ops);
+
+	/*
+	 * -ENODEV is not reported to the caller.  The module itself
+         * will use the timer mode sampling as fallback and this is
+         * always available.
+	 */
+	hwsampler_available = oprofile_hwsampler_init(ops) == 0;
+
+	return 0;
 #else
 	return -ENODEV;
 #endif
diff --git a/arch/s390/oprofile/op_counter.h b/arch/s390/oprofile/op_counter.h
new file mode 100644
index 000000000000..1a8d3ca09014
--- /dev/null
+++ b/arch/s390/oprofile/op_counter.h
@@ -0,0 +1,23 @@
+/**
+ * arch/s390/oprofile/op_counter.h
+ *
+ *   Copyright (C) 2011 IBM Deutschland Entwicklung GmbH, IBM Corporation
+ *   Author(s): Andreas Krebbel (krebbel@linux.vnet.ibm.com)
+ *
+ * @remark Copyright 2011 OProfile authors
+ */
+
+#ifndef OP_COUNTER_H
+#define OP_COUNTER_H
+
+struct op_counter_config {
+	/* `enabled' maps to the hwsampler_file variable.  */
+	/* `count' maps to the oprofile_hw_interval variable.  */
+	/* `event' and `unit_mask' are unused. */
+	unsigned long kernel;
+	unsigned long user;
+};
+
+extern struct op_counter_config counter_config;
+
+#endif /* OP_COUNTER_H */
diff --git a/arch/x86/include/asm/insn.h b/arch/x86/include/asm/insn.h
index 88c765e16410..74df3f1eddfd 100644
--- a/arch/x86/include/asm/insn.h
+++ b/arch/x86/include/asm/insn.h
@@ -137,6 +137,13 @@ static inline int insn_is_avx(struct insn *insn)
 	return (insn->vex_prefix.value != 0);
 }
 
+/* Ensure this instruction is decoded completely */
+static inline int insn_complete(struct insn *insn)
+{
+	return insn->opcode.got && insn->modrm.got && insn->sib.got &&
+		insn->displacement.got && insn->immediate.got;
+}
+
 static inline insn_byte_t insn_vex_m_bits(struct insn *insn)
 {
 	if (insn->vex_prefix.nbytes == 2)	/* 2 bytes VEX */
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index f61c62f7d5d8..096c975e099f 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -57,6 +57,7 @@
 		(1 << (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX))
 
 #define ARCH_PERFMON_BRANCH_MISSES_RETIRED		6
+#define ARCH_PERFMON_EVENTS_COUNT			7
 
 /*
  * Intel "Architectural Performance Monitoring" CPUID
@@ -72,6 +73,19 @@ union cpuid10_eax {
 	unsigned int full;
 };
 
+union cpuid10_ebx {
+	struct {
+		unsigned int no_unhalted_core_cycles:1;
+		unsigned int no_instructions_retired:1;
+		unsigned int no_unhalted_reference_cycles:1;
+		unsigned int no_llc_reference:1;
+		unsigned int no_llc_misses:1;
+		unsigned int no_branch_instruction_retired:1;
+		unsigned int no_branch_misses_retired:1;
+	} split;
+	unsigned int full;
+};
+
 union cpuid10_edx {
 	struct {
 		unsigned int num_counters_fixed:5;
@@ -81,6 +95,15 @@ union cpuid10_edx {
 	unsigned int full;
 };
 
+struct x86_pmu_capability {
+	int		version;
+	int		num_counters_gp;
+	int		num_counters_fixed;
+	int		bit_width_gp;
+	int		bit_width_fixed;
+	unsigned int	events_mask;
+	int		events_mask_len;
+};
 
 /*
  * Fixed-purpose performance events:
@@ -89,23 +112,24 @@ union cpuid10_edx {
 /*
  * All 3 fixed-mode PMCs are configured via this single MSR:
  */
-#define MSR_ARCH_PERFMON_FIXED_CTR_CTRL			0x38d
+#define MSR_ARCH_PERFMON_FIXED_CTR_CTRL	0x38d
 
 /*
  * The counts are available in three separate MSRs:
  */
 
 /* Instr_Retired.Any: */
-#define MSR_ARCH_PERFMON_FIXED_CTR0			0x309
-#define X86_PMC_IDX_FIXED_INSTRUCTIONS			(X86_PMC_IDX_FIXED + 0)
+#define MSR_ARCH_PERFMON_FIXED_CTR0	0x309
+#define X86_PMC_IDX_FIXED_INSTRUCTIONS	(X86_PMC_IDX_FIXED + 0)
 
 /* CPU_CLK_Unhalted.Core: */
-#define MSR_ARCH_PERFMON_FIXED_CTR1			0x30a
-#define X86_PMC_IDX_FIXED_CPU_CYCLES			(X86_PMC_IDX_FIXED + 1)
+#define MSR_ARCH_PERFMON_FIXED_CTR1	0x30a
+#define X86_PMC_IDX_FIXED_CPU_CYCLES	(X86_PMC_IDX_FIXED + 1)
 
 /* CPU_CLK_Unhalted.Ref: */
-#define MSR_ARCH_PERFMON_FIXED_CTR2			0x30b
-#define X86_PMC_IDX_FIXED_BUS_CYCLES			(X86_PMC_IDX_FIXED + 2)
+#define MSR_ARCH_PERFMON_FIXED_CTR2	0x30b
+#define X86_PMC_IDX_FIXED_REF_CYCLES	(X86_PMC_IDX_FIXED + 2)
+#define X86_PMC_MSK_FIXED_REF_CYCLES	(1ULL << X86_PMC_IDX_FIXED_REF_CYCLES)
 
 /*
  * We model BTS tracing as another fixed-mode PMC.
@@ -202,6 +226,7 @@ struct perf_guest_switch_msr {
 };
 
 extern struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr);
+extern void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap);
 #else
 static inline perf_guest_switch_msr *perf_guest_get_msrs(int *nr)
 {
@@ -209,6 +234,11 @@ static inline perf_guest_switch_msr *perf_guest_get_msrs(int *nr)
 	return NULL;
 }
 
+static inline void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap)
+{
+	memset(cap, 0, sizeof(*cap));
+}
+
 static inline void perf_events_lapic_init(void)	{ }
 #endif
 
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 2bda212a0010..5adce1040b11 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -484,18 +484,195 @@ static inline int is_x86_event(struct perf_event *event)
 	return event->pmu == &pmu;
 }
 
+/*
+ * Event scheduler state:
+ *
+ * Assign events iterating over all events and counters, beginning
+ * with events with least weights first. Keep the current iterator
+ * state in struct sched_state.
+ */
+struct sched_state {
+	int	weight;
+	int	event;		/* event index */
+	int	counter;	/* counter index */
+	int	unassigned;	/* number of events to be assigned left */
+	unsigned long used[BITS_TO_LONGS(X86_PMC_IDX_MAX)];
+};
+
+/* Total max is X86_PMC_IDX_MAX, but we are O(n!) limited */
+#define	SCHED_STATES_MAX	2
+
+struct perf_sched {
+	int			max_weight;
+	int			max_events;
+	struct event_constraint	**constraints;
+	struct sched_state	state;
+	int			saved_states;
+	struct sched_state	saved[SCHED_STATES_MAX];
+};
+
+/*
+ * Initialize interator that runs through all events and counters.
+ */
+static void perf_sched_init(struct perf_sched *sched, struct event_constraint **c,
+			    int num, int wmin, int wmax)
+{
+	int idx;
+
+	memset(sched, 0, sizeof(*sched));
+	sched->max_events	= num;
+	sched->max_weight	= wmax;
+	sched->constraints	= c;
+
+	for (idx = 0; idx < num; idx++) {
+		if (c[idx]->weight == wmin)
+			break;
+	}
+
+	sched->state.event	= idx;		/* start with min weight */
+	sched->state.weight	= wmin;
+	sched->state.unassigned	= num;
+}
+
+static void perf_sched_save_state(struct perf_sched *sched)
+{
+	if (WARN_ON_ONCE(sched->saved_states >= SCHED_STATES_MAX))
+		return;
+
+	sched->saved[sched->saved_states] = sched->state;
+	sched->saved_states++;
+}
+
+static bool perf_sched_restore_state(struct perf_sched *sched)
+{
+	if (!sched->saved_states)
+		return false;
+
+	sched->saved_states--;
+	sched->state = sched->saved[sched->saved_states];
+
+	/* continue with next counter: */
+	clear_bit(sched->state.counter++, sched->state.used);
+
+	return true;
+}
+
+/*
+ * Select a counter for the current event to schedule. Return true on
+ * success.
+ */
+static bool __perf_sched_find_counter(struct perf_sched *sched)
+{
+	struct event_constraint *c;
+	int idx;
+
+	if (!sched->state.unassigned)
+		return false;
+
+	if (sched->state.event >= sched->max_events)
+		return false;
+
+	c = sched->constraints[sched->state.event];
+
+	/* Prefer fixed purpose counters */
+	if (x86_pmu.num_counters_fixed) {
+		idx = X86_PMC_IDX_FIXED;
+		for_each_set_bit_cont(idx, c->idxmsk, X86_PMC_IDX_MAX) {
+			if (!__test_and_set_bit(idx, sched->state.used))
+				goto done;
+		}
+	}
+	/* Grab the first unused counter starting with idx */
+	idx = sched->state.counter;
+	for_each_set_bit_cont(idx, c->idxmsk, X86_PMC_IDX_FIXED) {
+		if (!__test_and_set_bit(idx, sched->state.used))
+			goto done;
+	}
+
+	return false;
+
+done:
+	sched->state.counter = idx;
+
+	if (c->overlap)
+		perf_sched_save_state(sched);
+
+	return true;
+}
+
+static bool perf_sched_find_counter(struct perf_sched *sched)
+{
+	while (!__perf_sched_find_counter(sched)) {
+		if (!perf_sched_restore_state(sched))
+			return false;
+	}
+
+	return true;
+}
+
+/*
+ * Go through all unassigned events and find the next one to schedule.
+ * Take events with the least weight first. Return true on success.
+ */
+static bool perf_sched_next_event(struct perf_sched *sched)
+{
+	struct event_constraint *c;
+
+	if (!sched->state.unassigned || !--sched->state.unassigned)
+		return false;
+
+	do {
+		/* next event */
+		sched->state.event++;
+		if (sched->state.event >= sched->max_events) {
+			/* next weight */
+			sched->state.event = 0;
+			sched->state.weight++;
+			if (sched->state.weight > sched->max_weight)
+				return false;
+		}
+		c = sched->constraints[sched->state.event];
+	} while (c->weight != sched->state.weight);
+
+	sched->state.counter = 0;	/* start with first counter */
+
+	return true;
+}
+
+/*
+ * Assign a counter for each event.
+ */
+static int perf_assign_events(struct event_constraint **constraints, int n,
+			      int wmin, int wmax, int *assign)
+{
+	struct perf_sched sched;
+
+	perf_sched_init(&sched, constraints, n, wmin, wmax);
+
+	do {
+		if (!perf_sched_find_counter(&sched))
+			break;	/* failed */
+		if (assign)
+			assign[sched.state.event] = sched.state.counter;
+	} while (perf_sched_next_event(&sched));
+
+	return sched.state.unassigned;
+}
+
 int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign)
 {
 	struct event_constraint *c, *constraints[X86_PMC_IDX_MAX];
 	unsigned long used_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)];
-	int i, j, w, wmax, num = 0;
+	int i, wmin, wmax, num = 0;
 	struct hw_perf_event *hwc;
 
 	bitmap_zero(used_mask, X86_PMC_IDX_MAX);
 
-	for (i = 0; i < n; i++) {
+	for (i = 0, wmin = X86_PMC_IDX_MAX, wmax = 0; i < n; i++) {
 		c = x86_pmu.get_event_constraints(cpuc, cpuc->event_list[i]);
 		constraints[i] = c;
+		wmin = min(wmin, c->weight);
+		wmax = max(wmax, c->weight);
 	}
 
 	/*
@@ -521,60 +698,12 @@ int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign)
 		if (assign)
 			assign[i] = hwc->idx;
 	}
-	if (i == n)
-		goto done;
-
-	/*
-	 * begin slow path
-	 */
-
-	bitmap_zero(used_mask, X86_PMC_IDX_MAX);
 
-	/*
-	 * weight = number of possible counters
-	 *
-	 * 1    = most constrained, only works on one counter
-	 * wmax = least constrained, works on any counter
-	 *
-	 * assign events to counters starting with most
-	 * constrained events.
-	 */
-	wmax = x86_pmu.num_counters;
+	/* slow path */
+	if (i != n)
+		num = perf_assign_events(constraints, n, wmin, wmax, assign);
 
 	/*
-	 * when fixed event counters are present,
-	 * wmax is incremented by 1 to account
-	 * for one more choice
-	 */
-	if (x86_pmu.num_counters_fixed)
-		wmax++;
-
-	for (w = 1, num = n; num && w <= wmax; w++) {
-		/* for each event */
-		for (i = 0; num && i < n; i++) {
-			c = constraints[i];
-			hwc = &cpuc->event_list[i]->hw;
-
-			if (c->weight != w)
-				continue;
-
-			for_each_set_bit(j, c->idxmsk, X86_PMC_IDX_MAX) {
-				if (!test_bit(j, used_mask))
-					break;
-			}
-
-			if (j == X86_PMC_IDX_MAX)
-				break;
-
-			__set_bit(j, used_mask);
-
-			if (assign)
-				assign[i] = j;
-			num--;
-		}
-	}
-done:
-	/*
 	 * scheduling failed or is just a simulation,
 	 * free resources if necessary
 	 */
@@ -1119,6 +1248,7 @@ static void __init pmu_check_apic(void)
 
 static int __init init_hw_perf_events(void)
 {
+	struct x86_pmu_quirk *quirk;
 	struct event_constraint *c;
 	int err;
 
@@ -1147,8 +1277,8 @@ static int __init init_hw_perf_events(void)
 
 	pr_cont("%s PMU driver.\n", x86_pmu.name);
 
-	if (x86_pmu.quirks)
-		x86_pmu.quirks();
+	for (quirk = x86_pmu.quirks; quirk; quirk = quirk->next)
+		quirk->func();
 
 	if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) {
 		WARN(1, KERN_ERR "hw perf events %d > max(%d), clipping!",
@@ -1171,12 +1301,18 @@ static int __init init_hw_perf_events(void)
 
 	unconstrained = (struct event_constraint)
 		__EVENT_CONSTRAINT(0, (1ULL << x86_pmu.num_counters) - 1,
-				   0, x86_pmu.num_counters);
+				   0, x86_pmu.num_counters, 0);
 
 	if (x86_pmu.event_constraints) {
+		/*
+		 * event on fixed counter2 (REF_CYCLES) only works on this
+		 * counter, so do not extend mask to generic counters
+		 */
 		for_each_event_constraint(c, x86_pmu.event_constraints) {
-			if (c->cmask != X86_RAW_EVENT_