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

* 'perfcounters-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (574 commits) perf_counter: Turn off by default perf_counter: Add counter->id to the throttle event perf_counter: Better align code perf_counter: Rename L2 to LL cache perf_counter: Standardize event names perf_counter: Rename enums perf_counter tools: Clean up u64 usage perf_counter: Rename perf_counter_limit sysctl perf_counter: More paranoia settings perf_counter: powerpc: Implement generalized cache events for POWER processors perf_counters: powerpc: Add support for POWER7 processors perf_counter: Accurate period data perf_counter: Introduce struct for sample data perf_counter tools: Normalize data using per sample period data perf_counter: Annotate exit ctx recursion perf_counter tools: Propagate signals properly perf_counter tools: Small frequency related fixes perf_counter: More aggressive frequency adjustment perf_counter/x86: Fix the model number of Intel Core2 processors perf_counter, x86: Correct some event and umask values for Intel processors ...
author: Linus Torvalds <torvalds@linux-foundation.org> 2009-06-11 14:01:07 -0700
committer: Linus Torvalds <torvalds@linux-foundation.org> 2009-06-11 14:01:07 -0700
commit: 8a1ca8cedd108c8e76a6ab34079d0bbb4f244799 (patch)
tree: 636c715524f1718599209cc289908ea44b6cb859 /arch
parent: b640f042faa2a2fad6464f259a8afec06e2f6386 (diff)
parent: 940010c5a314a7bd9b498593bc6ba1718ac5aec5 (diff)
46 files changed, 6814 insertions, 72 deletions
diff --git a/arch/powerpc/include/asm/hw_irq.h b/arch/powerpc/include/asm/hw_irq.h
index b7e034b0a6d..20a44d0c9fd 100644
--- a/arch/powerpc/include/asm/hw_irq.h
+++ b/arch/powerpc/include/asm/hw_irq.h
@@ -131,5 +131,44 @@ static inline int irqs_disabled_flags(unsigned long flags)
  */
 struct irq_chip;
 
+#ifdef CONFIG_PERF_COUNTERS
+static inline unsigned long test_perf_counter_pending(void)
+{
+	unsigned long x;
+
+	asm volatile("lbz %0,%1(13)"
+		: "=r" (x)
+		: "i" (offsetof(struct paca_struct, perf_counter_pending)));
+	return x;
+}
+
+static inline void set_perf_counter_pending(void)
+{
+	asm volatile("stb %0,%1(13)" : :
+		"r" (1),
+		"i" (offsetof(struct paca_struct, perf_counter_pending)));
+}
+
+static inline void clear_perf_counter_pending(void)
+{
+	asm volatile("stb %0,%1(13)" : :
+		"r" (0),
+		"i" (offsetof(struct paca_struct, perf_counter_pending)));
+}
+
+extern void perf_counter_do_pending(void);
+
+#else
+
+static inline unsigned long test_perf_counter_pending(void)
+{
+	return 0;
+}
+
+static inline void set_perf_counter_pending(void) {}
+static inline void clear_perf_counter_pending(void) {}
+static inline void perf_counter_do_pending(void) {}
+#endif /* CONFIG_PERF_COUNTERS */
+
 #endif	/* __KERNEL__ */
 #endif	/* _ASM_POWERPC_HW_IRQ_H */
diff --git a/arch/powerpc/include/asm/paca.h b/arch/powerpc/include/asm/paca.h
index 082b3aedf14..6ef05572301 100644
--- a/arch/powerpc/include/asm/paca.h
+++ b/arch/powerpc/include/asm/paca.h
@@ -99,6 +99,7 @@ struct paca_struct {
 	u8 soft_enabled;		/* irq soft-enable flag */
 	u8 hard_enabled;		/* set if irqs are enabled in MSR */
 	u8 io_sync;			/* writel() needs spin_unlock sync */
+	u8 perf_counter_pending;	/* PM interrupt while soft-disabled */
 
 	/* Stuff for accurate time accounting */
 	u64 user_time;			/* accumulated usermode TB ticks */
diff --git a/arch/powerpc/include/asm/perf_counter.h b/arch/powerpc/include/asm/perf_counter.h
new file mode 100644
index 00000000000..cc7c887705b
--- /dev/null
+++ b/arch/powerpc/include/asm/perf_counter.h
@@ -0,0 +1,98 @@
+/*
+ * Performance counter support - PowerPC-specific definitions.
+ *
+ * Copyright 2008-2009 Paul Mackerras, IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+#include <linux/types.h>
+
+#define MAX_HWCOUNTERS		8
+#define MAX_EVENT_ALTERNATIVES	8
+#define MAX_LIMITED_HWCOUNTERS	2
+
+/*
+ * This struct provides the constants and functions needed to
+ * describe the PMU on a particular POWER-family CPU.
+ */
+struct power_pmu {
+	int	n_counter;
+	int	max_alternatives;
+	u64	add_fields;
+	u64	test_adder;
+	int	(*compute_mmcr)(u64 events[], int n_ev,
+				unsigned int hwc[], u64 mmcr[]);
+	int	(*get_constraint)(u64 event, u64 *mskp, u64 *valp);
+	int	(*get_alternatives)(u64 event, unsigned int flags,
+				    u64 alt[]);
+	void	(*disable_pmc)(unsigned int pmc, u64 mmcr[]);
+	int	(*limited_pmc_event)(u64 event);
+	u32	flags;
+	int	n_generic;
+	int	*generic_events;
+	int	(*cache_events)[PERF_COUNT_HW_CACHE_MAX]
+			       [PERF_COUNT_HW_CACHE_OP_MAX]
+			       [PERF_COUNT_HW_CACHE_RESULT_MAX];
+};
+
+extern struct power_pmu *ppmu;
+
+/*
+ * Values for power_pmu.flags
+ */
+#define PPMU_LIMITED_PMC5_6	1	/* PMC5/6 have limited function */
+#define PPMU_ALT_SIPR		2	/* uses alternate posn for SIPR/HV */
+
+/*
+ * Values for flags to get_alternatives()
+ */
+#define PPMU_LIMITED_PMC_OK	1	/* can put this on a limited PMC */
+#define PPMU_LIMITED_PMC_REQD	2	/* have to put this on a limited PMC */
+#define PPMU_ONLY_COUNT_RUN	4	/* only counting in run state */
+
+struct pt_regs;
+extern unsigned long perf_misc_flags(struct pt_regs *regs);
+#define perf_misc_flags(regs)	perf_misc_flags(regs)
+
+extern unsigned long perf_instruction_pointer(struct pt_regs *regs);
+
+/*
+ * The power_pmu.get_constraint function returns a 64-bit value and
+ * a 64-bit mask that express the constraints between this event and
+ * other events.
+ *
+ * The value and mask are divided up into (non-overlapping) bitfields
+ * of three different types:
+ *
+ * Select field: this expresses the constraint that some set of bits
+ * in MMCR* needs to be set to a specific value for this event.  For a
+ * select field, the mask contains 1s in every bit of the field, and
+ * the value contains a unique value for each possible setting of the
+ * MMCR* bits.  The constraint checking code will ensure that two events
+ * that set the same field in their masks have the same value in their
+ * value dwords.
+ *
+ * Add field: this expresses the constraint that there can be at most
+ * N events in a particular class.  A field of k bits can be used for
+ * N <= 2^(k-1) - 1.  The mask has the most significant bit of the field
+ * set (and the other bits 0), and the value has only the least significant
+ * bit of the field set.  In addition, the 'add_fields' and 'test_adder'
+ * in the struct power_pmu for this processor come into play.  The
+ * add_fields value contains 1 in the LSB of the field, and the
+ * test_adder contains 2^(k-1) - 1 - N in the field.
+ *
+ * NAND field: this expresses the constraint that you may not have events
+ * in all of a set of classes.  (For example, on PPC970, you can't select
+ * events from the FPU, ISU and IDU simultaneously, although any two are
+ * possible.)  For N classes, the field is N+1 bits wide, and each class
+ * is assigned one bit from the least-significant N bits.  The mask has
+ * only the most-significant bit set, and the value has only the bit
+ * for the event's class set.  The test_adder has the least significant
+ * bit set in the field.
+ *
+ * If an event is not subject to the constraint expressed by a particular
+ * field, then it will have 0 in both the mask and value for that field.
+ */
diff --git a/arch/powerpc/include/asm/reg.h b/arch/powerpc/include/asm/reg.h
index e8018d540e8..fb359b0a693 100644
--- a/arch/powerpc/include/asm/reg.h
+++ b/arch/powerpc/include/asm/reg.h
@@ -492,11 +492,13 @@
 #define   MMCR0_FCHV	0x00000001UL /* freeze conditions in hypervisor mode */
 #define SPRN_MMCR1	798
 #define SPRN_MMCRA	0x312
+#define   MMCRA_SDSYNC	0x80000000UL /* SDAR synced with SIAR */
 #define   MMCRA_SIHV	0x10000000UL /* state of MSR HV when SIAR set */
 #define   MMCRA_SIPR	0x08000000UL /* state of MSR PR when SIAR set */
 #define   MMCRA_SLOT	0x07000000UL /* SLOT bits (37-39) */
 #define   MMCRA_SLOT_SHIFT	24
 #define   MMCRA_SAMPLE_ENABLE 0x00000001UL /* enable sampling */
+#define   POWER6_MMCRA_SDSYNC 0x0000080000000000ULL	/* SDAR/SIAR synced */
 #define   POWER6_MMCRA_SIHV   0x0000040000000000ULL
 #define   POWER6_MMCRA_SIPR   0x0000020000000000ULL
 #define   POWER6_MMCRA_THRM	0x00000020UL
diff --git a/arch/powerpc/include/asm/systbl.h b/arch/powerpc/include/asm/systbl.h
index d98a30dfd41..a0b92de51c7 100644
--- a/arch/powerpc/include/asm/systbl.h
+++ b/arch/powerpc/include/asm/systbl.h
@@ -322,6 +322,6 @@ SYSCALL_SPU(epoll_create1)
 SYSCALL_SPU(dup3)
 SYSCALL_SPU(pipe2)
 SYSCALL(inotify_init1)
-SYSCALL(ni_syscall)
+SYSCALL_SPU(perf_counter_open)
 COMPAT_SYS_SPU(preadv)
 COMPAT_SYS_SPU(pwritev)
diff --git a/arch/powerpc/include/asm/unistd.h b/arch/powerpc/include/asm/unistd.h
index 3f06f8ec81c..4badac2d11d 100644
--- a/arch/powerpc/include/asm/unistd.h
+++ b/arch/powerpc/include/asm/unistd.h
@@ -341,6 +341,7 @@
 #define __NR_dup3		316
 #define __NR_pipe2		317
 #define __NR_inotify_init1	318
+#define __NR_perf_counter_open	319
 #define __NR_preadv		320
 #define __NR_pwritev		321
 
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index 71901fbda4a..a2c683403c2 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -94,6 +94,9 @@ obj64-$(CONFIG_AUDIT)		+= compat_audit.o
 
 obj-$(CONFIG_DYNAMIC_FTRACE)	+= ftrace.o
 obj-$(CONFIG_FUNCTION_GRAPH_TRACER)	+= ftrace.o
+obj-$(CONFIG_PERF_COUNTERS)	+= perf_counter.o power4-pmu.o ppc970-pmu.o \
+				   power5-pmu.o power5+-pmu.o power6-pmu.o \
+				   power7-pmu.o
 
 obj-$(CONFIG_8XX_MINIMAL_FPEMU) += softemu8xx.o
 
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index 1e40bc05394..e981d1ce191 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -131,6 +131,7 @@ int main(void)
 	DEFINE(PACAKMSR, offsetof(struct paca_struct, kernel_msr));
 	DEFINE(PACASOFTIRQEN, offsetof(struct paca_struct, soft_enabled));
 	DEFINE(PACAHARDIRQEN, offsetof(struct paca_struct, hard_enabled));
+	DEFINE(PACAPERFPEND, offsetof(struct paca_struct, perf_counter_pending));
 	DEFINE(PACASLBCACHE, offsetof(struct paca_struct, slb_cache));
 	DEFINE(PACASLBCACHEPTR, offsetof(struct paca_struct, slb_cache_ptr));
 	DEFINE(PACACONTEXTID, offsetof(struct paca_struct, context.id));
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index abfc3233047..43e073477c3 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -526,6 +526,15 @@ ALT_FW_FTR_SECTION_END_IFCLR(FW_FEATURE_ISERIES)
 2:
 	TRACE_AND_RESTORE_IRQ(r5);
 
+#ifdef CONFIG_PERF_COUNTERS
+	/* check paca->perf_counter_pending if we're enabling ints */
+	lbz	r3,PACAPERFPEND(r13)
+	and.	r3,r3,r5
+	beq	27f
+	bl	.perf_counter_do_pending
+27:
+#endif /* CONFIG_PERF_COUNTERS */
+
 	/* extract EE bit and use it to restore paca->hard_enabled */
 	ld	r3,_MSR(r1)
 	rldicl	r4,r3,49,63		/* r0 = (r3 >> 15) & 1 */
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index 8c1a4966867..feff792ed0f 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -135,6 +135,11 @@ notrace void raw_local_irq_restore(unsigned long en)
 			iseries_handle_interrupts();
 	}
 
+	if (test_perf_counter_pending()) {
+		clear_perf_counter_pending();
+		perf_counter_do_pending();
+	}
+
 	/*
 	 * if (get_paca()->hard_enabled) return;
 	 * But again we need to take care that gcc gets hard_enabled directly
diff --git a/arch/powerpc/kernel/perf_counter.c b/arch/powerpc/kernel/perf_counter.c
new file mode 100644
index 00000000000..bb202388170
--- /dev/null
+++ b/arch/powerpc/kernel/perf_counter.c
@@ -0,0 +1,1263 @@
+/*
+ * Performance counter support - powerpc architecture code
+ *
+ * Copyright 2008-2009 Paul Mackerras, IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/perf_counter.h>
+#include <linux/percpu.h>
+#include <linux/hardirq.h>
+#include <asm/reg.h>
+#include <asm/pmc.h>
+#include <asm/machdep.h>
+#include <asm/firmware.h>
+#include <asm/ptrace.h>
+
+struct cpu_hw_counters {
+	int n_counters;
+	int n_percpu;
+	int disabled;
+	int n_added;
+	int n_limited;
+	u8  pmcs_enabled;
+	struct perf_counter *counter[MAX_HWCOUNTERS];
+	u64 events[MAX_HWCOUNTERS];
+	unsigned int flags[MAX_HWCOUNTERS];
+	u64 mmcr[3];
+	struct perf_counter *limited_counter[MAX_LIMITED_HWCOUNTERS];
+	u8  limited_hwidx[MAX_LIMITED_HWCOUNTERS];
+};
+DEFINE_PER_CPU(struct cpu_hw_counters, cpu_hw_counters);
+
+struct power_pmu *ppmu;
+
+/*
+ * Normally, to ignore kernel events we set the FCS (freeze counters
+ * in supervisor mode) bit in MMCR0, but if the kernel runs with the
+ * hypervisor bit set in the MSR, or if we are running on a processor
+ * where the hypervisor bit is forced to 1 (as on Apple G5 processors),
+ * then we need to use the FCHV bit to ignore kernel events.
+ */
+static unsigned int freeze_counters_kernel = MMCR0_FCS;
+
+static void perf_counter_interrupt(struct pt_regs *regs);
+
+void perf_counter_print_debug(void)
+{
+}
+
+/*
+ * Read one performance monitor counter (PMC).
+ */
+static unsigned long read_pmc(int idx)
+{
+	unsigned long val;
+
+	switch (idx) {
+	case 1:
+		val = mfspr(SPRN_PMC1);
+		break;
+	case 2:
+		val = mfspr(SPRN_PMC2);
+		break;
+	case 3:
+		val = mfspr(SPRN_PMC3);
+		break;
+	case 4:
+		val = mfspr(SPRN_PMC4);
+		break;
+	case 5:
+		val = mfspr(SPRN_PMC5);
+		break;
+	case 6:
+		val = mfspr(SPRN_PMC6);
+		break;
+	case 7:
+		val = mfspr(SPRN_PMC7);
+		break;
+	case 8:
+		val = mfspr(SPRN_PMC8);
+		break;
+	default:
+		printk(KERN_ERR "oops trying to read PMC%d\n", idx);
+		val = 0;
+	}
+	return val;
+}
+
+/*
+ * Write one PMC.
+ */
+static void write_pmc(int idx, unsigned long val)
+{
+	switch (idx) {
+	case 1:
+		mtspr(SPRN_PMC1, val);
+		break;
+	case 2:
+		mtspr(SPRN_PMC2, val);
+		break;
+	case 3:
+		mtspr(SPRN_PMC3, val);
+		break;
+	case 4:
+		mtspr(SPRN_PMC4, val);
+		break;
+	case 5:
+		mtspr(SPRN_PMC5, val);
+		break;
+	case 6:
+		mtspr(SPRN_PMC6, val);
+		break;
+	case 7:
+		mtspr(SPRN_PMC7, val);
+		break;
+	case 8:
+		mtspr(SPRN_PMC8, val);
+		break;
+	default:
+		printk(KERN_ERR "oops trying to write PMC%d\n", idx);
+	}
+}
+
+/*
+ * Check if a set of events can all go on the PMU at once.
+ * If they can't, this will look at alternative codes for the events
+ * and see if any combination of alternative codes is feasible.
+ * The feasible set is returned in event[].
+ */
+static int power_check_constraints(u64 event[], unsigned int cflags[],
+				   int n_ev)
+{
+	u64 mask, value, nv;
+	u64 alternatives[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES];
+	u64 amasks[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES];
+	u64 avalues[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES];
+	u64 smasks[MAX_HWCOUNTERS], svalues[MAX_HWCOUNTERS];
+	int n_alt[MAX_HWCOUNTERS], choice[MAX_HWCOUNTERS];
+	int i, j;
+	u64 addf = ppmu->add_fields;
+	u64 tadd = ppmu->test_adder;
+
+	if (n_ev > ppmu->n_counter)
+		return -1;
+
+	/* First see if the events will go on as-is */
+	for (i = 0; i < n_ev; ++i) {
+		if ((cflags[i] & PPMU_LIMITED_PMC_REQD)
+		    && !ppmu->limited_pmc_event(event[i])) {
+			ppmu->get_alternatives(event[i], cflags[i],
+					       alternatives[i]);
+			event[i] = alternatives[i][0];
+		}
+		if (ppmu->get_constraint(event[i], &amasks[i][0],
+					 &avalues[i][0]))
+			return -1;
+	}
+	value = mask = 0;
+	for (i = 0; i < n_ev; ++i) {
+		nv = (value | avalues[i][0]) + (value & avalues[i][0] & addf);
+		if ((((nv + tadd) ^ value) & mask) != 0 ||
+		    (((nv + tadd) ^ avalues[i][0]) & amasks[i][0]) != 0)
+			break;
+		value = nv;
+		mask |= amasks[i][0];
+	}
+	if (i == n_ev)
+		return 0;	/* all OK */
+
+	/* doesn't work, gather alternatives... */
+	if (!ppmu->get_alternatives)
+		return -1;
+	for (i = 0; i < n_ev; ++i) {
+		choice[i] = 0;
+		n_alt[i] = ppmu->get_alternatives(event[i], cflags[i],
+						  alternatives[i]);
+		for (j = 1; j < n_alt[i]; ++j)
+			ppmu->get_constraint(alternatives[i][j],
+					     &amasks[i][j], &avalues[i][j]);
+	}
+
+	/* enumerate all possibilities and see if any will work */
+	i = 0;
+	j = -1;
+	value = mask = nv = 0;
+	while (i < n_ev) {
+		if (j >= 0) {
+			/* we're backtracking, restore context */
+			value = svalues[i];
+			mask = smasks[i];
+			j = choice[i];
+		}
+		/*
+		 * See if any alternative k for event i,
+		 * where k > j, will satisfy the constraints.
+		 */
+		while (++j < n_alt[i]) {
+			nv = (value | avalues[i][j]) +
+				(value & avalues[i][j] & addf);
+			if ((((nv + tadd) ^ value) & mask) == 0 &&
+			    (((nv + tadd) ^ avalues[i][j])
+			     & amasks[i][j]) == 0)
+				break;
+		}
+		if (j >= n_alt[i]) {
+			/*
+			 * No feasible alternative, backtrack
+			 * to event i-1 and continue enumerating its
+			 * alternatives from where we got up to.
+			 */
+			if (--i < 0)
+				return -1;
+		} else {
+			/*
+			 * Found a feasible alternative for event i,
+			 * remember where we got up to with this event,
+			 * go on to the next event, and start with
+			 * the first alternative for it.
+			 */
+			choice[i] = j;
+			svalues[i] = value;
+			smasks[i] = mask;
+			value = nv;
+			mask |= amasks[i][j];
+			++i;
+			j = -1;
+		}
+	}
+
+	/* OK, we have a feasible combination, tell the caller the solution */
+	for (i = 0; i < n_ev; ++i)
+		event[i] = alternatives[i][choice[i]];
+	return 0;
+}
+
+/*
+ * Check if newly-added counters have consistent settings for
+ * exclude_{user,kernel,hv} with each other and any previously
+ * added counters.
+ */
+static int check_excludes(struct perf_counter **ctrs, unsigned int cflags[],
+			  int n_prev, int n_new)
+{
+	int eu = 0, ek = 0, eh = 0;
+	int i, n, first;
+	struct perf_counter *counter;
+
+	n = n_prev + n_new;
+	if (n <= 1)
+		return 0;
+
+	first = 1;
+	for (i = 0; i < n; ++i) {
+		if (cflags[i] & PPMU_LIMITED_PMC_OK) {
+			cflags[i] &= ~PPMU_LIMITED_PMC_REQD;
+			continue;
+		}
+		counter = ctrs[i];
+		if (first) {
+			eu = counter->attr.exclude_user;
+			ek = counter->attr.exclude_kernel;
+			eh = counter->attr.exclude_hv;
+			first = 0;
+		} else if (counter->attr.exclude_user != eu ||
+			   counter->attr.exclude_kernel != ek ||
+			   counter->attr.exclude_hv != eh) {
+			return -EAGAIN;
+		}
+	}
+
+	if (eu || ek || eh)
+		for (i = 0; i < n; ++i)
+			if (cflags[i] & PPMU_LIMITED_PMC_OK)
+				cflags[i] |= PPMU_LIMITED_PMC_REQD;
+
+	return 0;
+}
+
+static void power_pmu_read(struct perf_counter *counter)
+{
+	long val, delta, prev;
+
+	if (!counter->hw.idx)
+		return;
+	/*
+	 * Performance monitor interrupts come even when interrupts
+	 * are soft-disabled, as long as interrupts are hard-enabled.
+	 * Therefore we treat them like NMIs.
+	 */
+	do {
+		prev = atomic64_read(&counter->hw.prev_count);
+		barrier();
+		val = read_pmc(counter->hw.idx);
+	} while (atomic64_cmpxchg(&counter->hw.prev_count, prev, val) != prev);
+
+	/* The counters are only 32 bits wide */
+	delta = (val - prev) & 0xfffffffful;
+	atomic64_add(delta, &counter->count);
+	atomic64_sub(delta, &counter->hw.period_left);
+}
+
+/*
+ * On some machines, PMC5 and PMC6 can't be written, don't respect
+ * the freeze conditions, and don't generate interrupts.  This tells
+ * us if `counter' is using such a PMC.
+ */
+static int is_limited_pmc(int pmcnum)
+{
+	return (ppmu->flags & PPMU_LIMITED_PMC5_6)
+		&& (pmcnum == 5 || pmcnum == 6);
+}
+
+static void freeze_limited_counters(struct cpu_hw_counters *cpuhw,
+				    unsigned long pmc5, unsigned long pmc6)
+{
+	struct perf_counter *counter;
+	u64 val, prev, delta;
+	int i;
+
+	for (i = 0; i < cpuhw->n_limited; ++i) {
+		counter = cpuhw->limited_counter[i];
+		if (!counter->hw.idx)
+			continue;
+		val = (counter->hw.idx == 5) ? pmc5 : pmc6;
+		prev = atomic64_read(&counter->hw.prev_count);
+		counter->hw.idx = 0;
+		delta = (val - prev) & 0xfffffffful;
+		atomic64_add(delta, &counter->count);
+	}
+}
+
+static void thaw_limited_counters(struct cpu_hw_counters *cpuhw,
+				  unsigned long pmc5, unsigned long pmc6)
+{
+	struct perf_counter *counter;
+	u64 val;
+	int i;
+
+	for (i = 0; i < cpuhw->n_limited; ++i) {
+		counter = cpuhw->limited_counter[i];
+		counter->hw.idx = cpuhw->limited_hwidx[i];
+		val = (counter->hw.idx == 5) ? pmc5 : pmc6;
+		atomic64_set(&counter->hw.prev_count, val);
+		perf_counter_update_userpage(counter);
+	}
+}
+
+/*
+ * Since limited counters don't respect the freeze conditions, we
+ * have to read them immediately after freezing or unfreezing the
+ * other counters.  We try to keep the values from the limited
+ * counters as consistent as possible by keeping the delay (in
+ * cycles and instructions) between freezing/unfreezing and reading
+ * the limited counters as small and consistent as possible.
+ * Therefore, if any limited counters are in use, we read them
+ * both, and always in the same order, to minimize variability,
+ * and do it inside the same asm that writes MMCR0.
+ */
+static void write_mmcr0(struct cpu_hw_counters *cpuhw, unsigned long mmcr0)
+{
+	unsigned long pmc5, pmc6;
+
+	if (!cpuhw->n_limited) {
+		mtspr(SPRN_MMCR0, mmcr0);
+		return;
+	}
+
+	/*
+	 * Write MMCR0, then read PMC5 and PMC6 immediately.
+	 * To ensure we don't get a performance monitor interrupt
+	 * between writing MMCR0 and freezing/thawing the limited
+	 * counters, we first write MMCR0 with the counter overflow
+	 * interrupt enable bits turned off.
+	 */
+	asm volatile("mtspr %3,%2; mfspr %0,%4; mfspr %1,%5"
+		     : "=&r" (pmc5), "=&r" (pmc6)
+		     : "r" (mmcr0 & ~(MMCR0_PMC1CE | MMCR0_PMCjCE)),
+		       "i" (SPRN_MMCR0),
+		       "i" (SPRN_PMC5), "i" (SPRN_PMC6));
+
+	if (mmcr0 & MMCR0_FC)
+		freeze_limited_counters(cpuhw, pmc5, pmc6);
+	else
+		thaw_limited_counters(cpuhw, pmc5, pmc6);
+
+	/*
+	 * Write the full MMCR0 including the counter overflow interrupt
+	 * enable bits, if necessary.
+	 */
+	if (mmcr0 & (MMCR0_PMC1CE | MMCR0_PMCjCE))
+		mtspr(SPRN_MMCR0, mmcr0);
+}
+
+/*
+ * Disable all counters to prevent PMU interrupts and to allow
+ * counters to be added or removed.
+ */
+void hw_perf_disable(void)
+{
+	struct cpu_hw_counters *cpuhw;
+	unsigned long ret;
+	unsigned long flags;
+
+	local_irq_save(flags);
+	cpuhw = &__get_cpu_var(cpu_hw_counters);
+
+	ret = cpuhw->disabled;
+	if (!ret) {
+		cpuhw->disabled = 1;
+		cpuhw->n_added = 0;
+
+		/*
+		 * Check if we ever enabled the PMU on this cpu.
+		 */
+		if (!cpuhw->pmcs_enabled) {
+			if (ppc_md.enable_pmcs)
+				ppc_md.enable_pmcs();
+			cpuhw->pmcs_enabled = 1;
+		}
+
+		/*
+		 * Disable instruction sampling if it was enabled
+		 */
+		if (cpuhw->mmcr[2] & MMCRA_SAMPLE_ENABLE) {
+			mtspr(SPRN_MMCRA,
+			      cpuhw->mmcr[2] & ~MMCRA_SAMPLE_ENABLE);
+			mb();
+		}
+
+		/*
+		 * Set the 'freeze counters' bit.
+		 * The barrier is to make sure the mtspr has been
+		 * executed and the PMU has frozen the counters
+		 * before we return.
+		 */
+		write_mmcr0(cpuhw, mfspr(SPRN_MMCR0) | MMCR0_FC);
+		mb();
+	}
+	local_irq_restore(flags);
+}
+
+/*
+ * Re-enable all counters if disable == 0.
+ * If we were previously disabled and counters were added, then
+ * put the new config on the PMU.
+ */
+void hw_perf_enable(void)
+{
+	struct perf_counter *counter;
+	struct cpu_hw_counters *cpuhw;
+	unsigned long flags;
+	long i;
+	unsigned long val;
+	s64 left;
+	unsigned int hwc_index[MAX_HWCOUNTERS];
+	int n_lim;
+	int idx;
+
+	local_irq_save(flags);
+	cpuhw = &__get_cpu_var(cpu_hw_counters);
+	if (!cpuhw->disabled) {
+		local_irq_restore(flags);
+		return;
+	}
+	cpuhw->disabled = 0;
+
+	/*
+	 * If we didn't change anything, or only removed counters,
+	 * no need to recalculate MMCR* settings and reset the PMCs.
+	 * Just reenable the PMU with the current MMCR* settings
+	 * (possibly updated for removal of counters).
+	 */
+	if (!cpuhw->n_added) {
+		mtspr(SPRN_MMCRA, cpuhw->mmcr[2] & ~MMCRA_SAMPLE_ENABLE);
+		mtspr(SPRN_MMCR1, cpuhw->mmcr[1]);
+		if (cpuhw->n_counters == 0)
+			get_lppaca()->pmcregs_in_use = 0;
+		goto out_enable;
+	}
+
+	/*
+	 * Compute MMCR* values for the new set of counters
+	 */
+	if (ppmu->compute_mmcr(cpuhw->events, cpuhw->n_counters, hwc_index,
+			       cpuhw->mmcr)) {
+		/* shouldn't ever get here */
+		printk(KERN_ERR "oops compute_mmcr failed\n");
+		goto out;
+	}
+
+	/*
+	 * Add in MMCR0 freeze bits corresponding to the
+	 * attr.exclude_* bits for the first counter.
+	 * We have already checked that all counters have the
+	 * same values for these bits as the first counter.
+	 */
+	counter = cpuhw->counter[0];
+	if (counter->attr.exclude_user)
+		cpuhw->mmcr[0] |= MMCR0_FCP;
+	if (counter->attr.exclude_kernel)
+		cpuhw->mmcr[0] |= freeze_counters_kernel;
+	if (counter->attr.exclude_hv)
+		cpuhw->mmcr[0] |= MMCR0_FCHV;
+
+	/*
+	 * Write the new configuration to MMCR* with the freeze
+	 * bit set and set the hardware counters to their initial values.
+	 * Then unfreeze the counters.
+	 */
+	get_lppaca()->pmcregs_in_use = 1;
+	mtspr(SPRN_MMCRA, cpuhw->mmcr[2] & ~MMCRA_SAMPLE_ENABLE);
+	mtspr(SPRN_MMCR1, cpuhw->mmcr[1]);
+	mtspr(SPRN_MMCR0, (cpuhw->mmcr[0] & ~(MMCR0_PMC1CE | MMCR0_PMCjCE))
+				| MMCR0_FC);
+
+	/*
+	 * Read off any pre-existing counters that need to move
+	 * to another PMC.
+	 */
+	for (i = 0; i < cpuhw->n_counters; ++i) {
+		counter = cpuhw->counter[i];
+		if (counter->hw.idx && counter->hw.idx != hwc_index[i] + 1) {
+			power_pmu_read(counter);
+			write_pmc(counter->hw.idx, 0);
+			counter->hw.idx = 0;
+		}
+	}
+
+	/*
+	 * Initialize the PMCs for all the new and moved counters.
+	 */
+	cpuhw->n_limited = n_lim = 0;
+	for (i = 0; i < cpuhw->n_counters; ++i) {
+		counter = cpuhw->counter[i];
+		if (counter->hw.idx)
+			continue;
+		idx = hwc_index[i] + 1;
+		if (is_limited_pmc(idx)) {
+			cpuhw->limited_counter[n_lim] = counter;
+			cpuhw->limited_hwidx[n_lim] = idx;
+			++n_lim;
+			continue;
+		}
+		val = 0;
+		if (counter->hw.sample_period) {
+			left = atomic64_read(&counter->hw.period_left);
+			if (left < 0x80000000L)
+				val = 0x80000000L - left;
+		}
+		atomic64_set(&counter->hw.prev_count, val);
+		counter->hw.idx = idx;
+		write_pmc(idx, val);
+		perf_counter_update_userpage(counter);
+	}
+	cpuhw->n_limited = n_lim;
+	cpuhw->mmcr[0] |= MMCR0_PMXE | MMCR0_FCECE;
+
+ out_enable:
+	mb();
+	write_mmcr0(cpuhw, cpuhw->mmcr[0]);
+
+	/*
+	 * Enable instruction sampling if necessary
+	 */
+	if (cpuhw->mmcr[2] & MMCRA_SAMPLE_ENABLE) {
+		mb();
+		mtspr(SPRN_MMCRA, cpuhw->mmcr[2]);
+	}
+
+ out:
+	local_irq_restore(flags);
+}
+
+static int collect_events(struct perf_counter *group, int max_count,
+			  struct perf_counter *ctrs[], u64 *events,
+			  unsigned int *flags)
+{
+	int n = 0;
+	struct perf_counter *counter;
+
+	if (!is_software_counter(group)) {
+		if (n >= max_count)
+			return -1;
+		ctrs[n] = group;
+		flags[n] = group->hw.counter_base;
+		events[n++] = group->hw.config;
+	}
+	list_for_each_entry(counter, &group->sibling_list, list_entry) {
+		if (!is_software_counter(counter) &&
+		    counter->state != PERF_COUNTER_STATE_OFF) {
+			if (n >= max_count)
+				return -1;
+			ctrs[n] = counter;
+			flags[n] = counter->hw.counter_base;
+			events[n++] = counter->hw.config;
+		}
+	}
+	return n;
+}
+
+static void counter_sched_in(struct perf_counter *counter, int cpu)
+{
+	counter->state = PERF_COUNTER_STATE_ACTIVE;
+	counter->oncpu = cpu;
+	counter->tstamp_running += counter->ctx->time - counter->tstamp_stopped;
+	if (is_software_counter(counter))
+		counter->pmu->enable(counter);
+}
+
+/*
+ * Called to enable a whole group of counters.
+ * Returns 1 if the group was enabled, or -EAGAIN if it could not be.
+ * Assumes the caller has disabled interrupts and has
+ * frozen the PMU with hw_perf_save_disable.
+ */
+int hw_perf_group_sched_in(struct perf_counter *group_leader,
+	       struct perf_cpu_context *cpuctx,
+	       struct perf_counter_context *ctx, int cpu)
+{
+	struct cpu_hw_counters *cpuhw;
+	long i, n, n0;
+	struct perf_counter *sub;
+
+	cpuhw = &__get_cpu_var(cpu_hw_counters);
+	n0 = cpuhw->n_counters;
+	n = collect_events(group_leader, ppmu->n_counter - n0,
+			   &cpuhw->counter[n0], &cpuhw->events[n0],
+			   &cpuhw->flags[n0]);
+	if (n < 0)
+		return -EAGAIN;
+	if (check_excludes(cpuhw->counter, cpuhw->flags, n0, n))
+		return -EAGAIN;
+	i = power_check_constraints(cpuhw->events, cpuhw->flags, n + n0);
+	if (i < 0)
+		return -EAGAIN;
+	cpuhw->n_counters = n0 + n;
+	cpuhw->n_added += n;
+
+	/*
+	 * OK, this group can go on; update counter states etc.,
+	 * and enable any software counters
+	 */
+	for (i = n0; i < n0 + n; ++i)
+		cpuhw->counter[i]->hw.config = cpuhw->events[i];
+	cpuctx->active_oncpu += n;
+	n = 1;
+	counter_sched_in(group_leader, cpu);
+	list_for_each_entry(sub, &group_leader->sibling_list, list_entry) {
+		if (sub->state != PERF_COUNTER_STATE_OFF) {
+			counter_sched_in(sub, cpu);
+			++n;
+		}
+	}
+	ctx->nr_active += n;
+
+	return 1;
+}
+
+/*
+ * Add a counter to the PMU.
+ * If all counters are not already frozen, then we disable and
+ * re-enable the PMU in order to get hw_perf_enable to do the
+ * actual work of reconfiguring the PMU.
+ */
+static int power_pmu_enable(struct perf_counter *counter)
+{
+	struct cpu_hw_counters *cpuhw;
+	unsigned long flags;
+	int n0;
+	int ret = -EAGAIN;
+
+	local_irq_save(flags);
+	perf_disable();
+
+	/*
+	 * Add the counter to the list (if there is room)
+	 * and check whether the total set is still feasible.
+	 */
+	cpuhw = &__get_cpu_var(cpu_hw_counters);
+	n0 = cpuhw->n_counters;
+	if (n0 >= ppmu->n_counter)
+		goto out;
+	cpuhw->counter[n0] = counter;
+	cpuhw->events[n0] = counter->hw.config;
+	cpuhw->flags[n0] = counter->hw.counter_base;
+	if (check_excludes(cpuhw->counter, cpuhw->flags, n0, 1))
+		goto out;
+	if (power_check_constraints(cpuhw->events, cpuhw->flags, n0 + 1))
+		goto out;
+
+	counter->hw.config = cpuhw->events[n0];
+	++cpuhw->n_counters;
+	++cpuhw->n_added;
+
+	ret = 0;
+ out:
+	perf_enable();
+	local_irq_restore(flags);
+	return ret;
+}
+
+/*
+ * Remove a counter from the PMU.
+ */
+static void power_pmu_disable(struct perf_counter *counter)
+{
+	struct cpu_hw_counters *cpuhw;
+	long i;
+	unsigned long flags;
+
+	local_irq_save(flags);
+	perf_disable();
+
+	power_pmu_read(counter);
+
+	cpuhw = &__get_cpu_var(cpu_hw_counters);
+	for (i = 0; i < cpuhw->n_counters; ++i) {
+		if (counter == cpuhw->counter[i]) {
+			while (++i < cpuhw->n_counters)
+				cpuhw->counter[i-1] = cpuhw->counter[i];
+			--cpuhw->n_counters;
+			ppmu->disable_pmc(counter->hw.idx - 1, cpuhw->mmcr);
+			if (counter->hw.idx) {
+				write_pmc(counter->hw.idx, 0);
+				counter->hw.idx = 0;
+			}
+			perf_counter_update_userpage(counter);
+			break;
+		}
+	}
+	for (i = 0; i < cpuhw->n_limited; ++i)
+		if (counter == cpuhw->limited_counter[i])
+			break;
+	if (i < cpuhw->n_limited) {
+		while (++i < cpuhw->n_limited) {
+			cpuhw->limited_counter[i-1] = cpuhw->limited_counter[i];
+			cpuhw->limited_hwidx[i-1] = cpuhw->limited_hwidx[i];
+		}
+		--cpuhw->n_limited;
+	}
+	if (cpuhw->n_counters == 0) {
+		/* disable exceptions if no counters are running */
+		cpuhw->mmcr[0] &= ~(MMCR0_PMXE | MMCR0_FCECE);
+	}
+
+	perf_enable();
+	local_irq_restore(flags);
+}
+
+/*
+ * Re-enable interrupts on a counter after they were throttled
+ * because they were coming too fast.
+ */
+static void power_pmu_unthrottle(struct perf_counter *counter)
+{
+	s64 val, left;
+	unsigned long flags;
+
+	if (!counter->hw.idx || !counter->hw.sample_period)
+		return;
+	local_irq_save(flags);
+	perf_disable();
+	power_pmu_read(counter);
+	left = counter->hw.sample_period;
+	counter->hw.last_period = left;
+	val = 0;
+	if (left < 0x80000000L)
+		val = 0x80000000L - left;
+	write_pmc(counter->hw.idx, val);
+	atomic64_set(&counter->hw.prev_count, val);
+	atomic64_set(&counter->hw.period_left, left);
+	perf_counter_update_userpage(counter);
+	perf_enable();
+	local_irq_restore(flags);
+}
+
+struct pmu power_pmu = {
+	.enable		= power_pmu_enable,
+	.disable	= power_pmu_disable,
+	.read		= power_pmu_read,
+	.unthrottle	= power_pmu_unthrottle,
+};
+
+/*
+ * Return 1 if we might be able to put counter on a limited PMC,
+ * or 0 if not.
+ * A counter can only go on a limited PMC if it counts something
+ * that a limited PMC can count, doesn't require interrupts, and
+ * doesn't exclude any processor mode.
+ */
+static int can_go_on_limited_pmc(struct perf_counter *counter, u64 ev,
+				 unsigned int flags)
+{
+	int n;
+	u64 alt[MAX_EVENT_ALTERNATIVES];
+
+	if (counter->attr.exclude_user
+	    || counter->attr.exclude_kernel
+	    || counter->attr.exclude_hv
+	    || counter->attr.sample_period)
+		return 0;
+
+	if (ppmu->limited_pmc_event(ev))
+		return 1;
+
+	/*
+	 * The requested event isn't on a limited PMC already;
+	 * see if any alternative code goes on a limited PMC.
+	 */
+	if (!ppmu->get_alternatives)
+		return 0;
+
+	flags |= PPMU_LIMITED_PMC_OK | PPMU_LIMITED_PMC_REQD;
+	n = ppmu->get_alternatives(ev, flags, alt);
+
+	return n > 0;
+}
+
+/*
+ * Find an alternative event that goes on a normal PMC, if possible,
+ * and return the event code, or 0 if there is no such alternative.
+ * (Note: event code 0 is "don't count" on all machines.)
+ */
+static u64 normal_pmc_alternative(u64 ev, unsigned long flags)
+{
+	u64 alt[MAX_EVENT_ALTERNATIVES];
+	int n;
+
+	flags &= ~(PPMU_LIMITED_PMC_OK | PPMU_LIMITED_PMC_REQD);
+	n = ppmu->get_alternatives(ev, flags, alt);
+	if (!n)
+		return 0;
+	return alt[0];
+}
+
+/* Number of perf_counters counting hardware events */
+static atomic_t num_counters;
+/* Used to avoid races in calling reserve/release_pmc_hardware */
+static DEFINE_MUTEX(pmc_reserve_mutex);
+
+/*
+ * Release the PMU if this is the last perf_counter.
+ */
+static void hw_perf_counter_destroy(struct perf_counter *counter)
+{
+	if (!atomic_add_unless(&num_counters, -1, 1)) {
+		mutex_lock(&pmc_reserve_mutex);
+		if (atomic_dec_return(&num_counters) == 0)
+			release_pmc_hardware();
+		mutex_unlock(&pmc_reserve_mutex);
+	}
+}
+
+/*
+ * Translate a generic cache event config to a raw event code.
+ */
+static int hw_perf_cache_event(u64 config, u64 *eventp)
+{
+	unsigned long type, op, result;
+	int ev;
+
+	if (!ppmu->cache_events)
+		return -EINVAL;
+
+	/* unpack config */
+	type = config & 0xff;
+	op = (config >> 8) & 0xff;
+	result = (config >> 16) & 0xff;
+
+	if (type >= PERF_COUNT_HW_CACHE_MAX ||
+	    op >= PERF_COUNT_HW_CACHE_OP_MAX ||
+	    result >= PERF_COUNT_HW_CACHE_RESULT_MAX)
+		return -EINVAL;
+
+	ev = (*ppmu->cache_events)[type][op][result];
+	if (ev == 0)
+		return -EOPNOTSUPP;
+	if (ev == -1)
+		return -EINVAL;
+	*eventp = ev;
+	return 0;
+}
+
+const struct pmu *hw_perf_counter_init(struct perf_counter *counter)
+{
+	u64 ev;
+	unsigned long flags;
+	struct perf_counter *ctrs[MAX_HWCOUNTERS];
+	u64 events[MAX_HWCOUNTERS];
+	unsigned int cflags[MAX_HWCOUNTERS];
+	int n;
+	int err;
+
+	if (!ppmu)
+		return ERR_PTR(-ENXIO);
+	switch (counter->attr.type) {
+	case PERF_TYPE_HARDWARE:
+		ev = counter->attr.config;
+		if (ev >= ppmu->n_generic || ppmu->generic_events[ev] == 0)
+			return ERR_PTR(-EOPNOTSUPP);
+		ev = ppmu->generic_events[ev];
+		break;
+	case PERF_TYPE_HW_CACHE:
+		err = hw_perf_cache_event(counter->attr.config, &ev);
+		if (err)
+			return ERR_PTR(err);
+		break;
+	case PERF_TYPE_RAW:
+		ev = counter->attr.config;
+		break;
+	}
+	counter->hw.config_base = ev;
+	counter->hw.idx = 0;
+
+	/*
+	 * If we are not running on a hypervisor, force the
+	 * exclude_hv bit to 0 so that we don't care what
+	 * the user set it to.
+	 */
+	if (!firmware_has_feature(FW_FEATURE_LPAR))
+		counter->attr.exclude_hv = 0;
+
+	/*
+	 * If this is a per-task counter, then we can use
+	 * PM_RUN_* events interchangeably with their non RUN_*
+	 * equivalents, e.g. PM_RUN_CYC instead of PM_CYC.
+	 * XXX we should check if the task is an idle task.
+	 */
+	flags = 0;
+	if (counter->ctx->task)
+		flags |= PPMU_ONLY_COUNT_RUN;
+
+	/*
+	 * If this machine has limited counters, check whether this
+	 * event could go on a limited counter.
+	 */
+	if (ppmu->flags & PPMU_LIMITED_PMC5_6) {
+		if (can_go_on_limited_pmc(counter, ev, flags)) {
+			flags |= PPMU_LIMITED_PMC_OK;
+		} else if (ppmu->limited_pmc_event(ev)) {
+			/*
+			 * The requested event is on a limited PMC,
+			 * but we can't use a limited PMC; see if any
+			 * alternative goes on a normal PMC.
+			 */
+			ev = normal_pmc_alternative(ev, flags);
+			if (!ev)
+				return ERR_PTR(-EINVAL);
+		}
+	}
+
+	/*
+	 * If this is in a group, check if it can go on with all the
+	 * other hardware counters in the group.  We assume the counter
+	 * hasn't been linked into its leader's sibling list at this point.
+	 */
+	n = 0;
+	if (counter->group_leader != counter) {
+		n = collect_events(counter->group_leader, ppmu->n_counter - 1,
+				   ctrs, events, cflags);
+		if (n < 0)
+			return ERR_PTR(-EINVAL);
+	}
+	events[n] = ev;
+	ctrs[n] = counter;
+	cflags[n] = flags;
+	if (check_excludes(ctrs, cflags, n, 1))
+		return ERR_PTR(-EINVAL);
+	if (power_check_constraints(events, cflags, n + 1))
+		return ERR_PTR(-EINVAL);
+
+	counter->hw.config = events[n];
+	counter->hw.counter_base = cflags[n];
+	counter->hw.last_period = counter->hw.sample_period;
+	atomic64_set(&counter->hw.period_left, counter->hw.last_period);
+
+	/*
+	 * See if we need to reserve the PMU.
+	 * If no counters are currently in use, then we have to take a
+	 * mutex to ensure that we don't race with another task doing
+	 * reserve_pmc_hardware or release_pmc_hardware.
+	 */
+	err = 0;
+	if (!atomic_inc_not_zero(&num_counters)) {
+		mutex_lock(&pmc_reserve_mutex);
+		if (atomic_read(&num_counters) == 0 &&
+		    reserve_pmc_hardware(perf_counter_interrupt))
+			err = -EBUSY;
+		else
+			atomic_inc(&num_counters);
+		mutex_unlock(&pmc_reserve_mutex);
+	}
+	counter->destroy = hw_perf_counter_destroy;
+
+	if (err)
+		return ERR_PTR(err);
+	return &power_pmu;
+}
+
+/*
+ * A counter has overflowed; update its count and record
+ * things if requested.  Note that interrupts are hard-disabled
+ * here so there is no possibility of being interrupted.
+ */
+static void record_and_restart(struct perf_counter *counter, long val,
+			       struct pt_regs *regs, int nmi)
+{
+	u64 period = counter->hw.sample_period;
+	s64 prev, delta, left;
+	int record = 0;
+	u64 addr, mmcra, sdsync;
+
+	/* we don't have to worry about interrupts here */
+	prev = atomic64_read(&counter->hw.prev_count);
+	delta = (val - prev) & 0xfffffffful;
+	atomic64_add(delta, &counter->count);
+
+	/*
+	 * See if the total period for this counter has expired,
+	 * and update for the next period.
+	 */
+	val = 0;
+	left = atomic64_read(&counter->hw.period_left) - delta;
+	if (period) {
+		if (left <= 0) {
+			left += period;
+			if (left <= 0)
+				left = period;
+			record = 1;
+		}
+		if (left < 0x80000000L)
+			val = 0x80000000L - left;
+	}
+
+	/*
+	 * Finally record data if requested.
+	 */
+	if (record) {
+		struct perf_sample_data data = {
+			.regs	= regs,
+			.addr	= 0,
+			.period	= counter->hw.last_period,
+		};
+
+		if (counter->attr.sample_type & PERF_SAMPLE_ADDR) {
+			/*
+			 * The user wants a data address recorded.
+			 * If we're not doing instruction sampling,
+			 * give them the SDAR (sampled data address).
+			 * If we are doing instruction sampling, then only
+			 * give them the SDAR if it corresponds to the
+			 * instruction pointed to by SIAR; this is indicated
+			 * by the [POWER6_]MMCRA_SDSYNC bit in MMCRA.
+			 */
+			mmcra = regs->dsisr;
+			sdsync = (ppmu->flags & PPMU_ALT_SIPR) ?
+				POWER6_MMCRA_SDSYNC : MMCRA_SDSYNC;
+			if (!(mmcra & MMCRA_SAMPLE_ENABLE) || (mmcra & sdsync))
+				data.addr = mfspr(SPRN_SDAR);
+		}
+		if (perf_counter_overflow(counter, nmi, &data)) {
+			/*
+			 * Interrupts are coming too fast - throttle them
+			 * by setting the counter to 0, so it will be
+			 * at least 2^30 cycles until the next interrupt
+			 * (assuming each counter counts at most 2 counts
+			 * per cycle).
+			 */
+			val = 0;
+			left = ~0ULL >> 1;
+		}
+	}
+
+	write_pmc(counter->hw.idx, val);
+	atomic64_set(&counter->hw.prev_count, val);
+	atomic64_set(&counter->hw.period_left, left);
+	perf_counter_update_userpage(counter);
+}
+
+/*
+ * Called from generic code to get the misc flags (i.e. processor mode)
+ * for an event.
+ */
+unsigned long perf_misc_flags(struct pt_regs *regs)
+{
+	unsigned long mmcra;
+
+	if (TRAP(regs) != 0xf00) {
+		/* not a PMU interrupt */
+		return user_mode(regs) ? PERF_EVENT_MISC_USER :
+			PERF_EVENT_MISC_KERNEL;
+	}
+
+	mmcra = regs->dsisr;
+	if (ppmu->flags & PPMU_ALT_SIPR) {
+		if (mmcra & POWER6_MMCRA_SIHV)
+			return PERF_EVENT_MISC_HYPERVISOR;
+		return (mmcra & POWER6_MMCRA_SIPR) ? PERF_EVENT_MISC_USER :
+			PERF_EVENT_MISC_KERNEL;
+	}
+	if (mmcra & MMCRA_SIHV)
+		return PERF_EVENT_MISC_HYPERVISOR;
+	return (mmcra & MMCRA_SIPR) ? PERF_EVENT_MISC_USER :
+			PERF_EVENT_MISC_KERNEL;
+}
+
+/*
+ * Called from generic code to get the instruction pointer
+ * for an event.
+ */
+unsigned long perf_instruction_pointer(struct pt_regs *regs)
+{
+	unsigned long mmcra;
+	unsigned long ip;
+	unsigned long slot;
+
+	if (TRAP(regs) != 0xf00)
+		return regs->nip;	/* not a PMU interrupt */
+
+	ip = mfspr(SPRN_SIAR);
+	mmcra = regs->dsisr;
+	if ((mmcra & MMCRA_SAMPLE_ENABLE) && !(ppmu->flags & PPMU_ALT_SIPR)) {
+		slot = (mmcra & MMCRA_SLOT) >> MMCRA_SLOT_SHIFT;
+		if (slot > 1)
+			ip += 4 * (slot - 1);
+	}
+	return ip;
+}
+
+/*
+ * Performance monitor interrupt stuff
+ */
+static void perf_counter_interrupt(struct pt_regs *regs)
+{
+	int i;
+	struct cpu_hw_counters *cpuhw = &__get_cpu_var(cpu_hw_counters);
+	struct perf_counter *counter;
+	long val;
+	int found = 0;
+	int nmi;
+
+	if (cpuhw->n_limited)
+		freeze_limited_counters(cpuhw, mfspr(SPRN_PMC5),
+					mfspr(SPRN_PMC6));
+
+	/*
+	 * Overload regs->dsisr to store MMCRA so we only need to read it once.
+	 */
+	regs->dsisr = mfspr(SPRN_MMCRA);
+
+	/*
+	 * If interrupts were soft-disabled when this PMU interrupt
+	 * occurred, treat it as an NMI.
+	 */
+	nmi = !regs->softe;
+	if (nmi)
+		nmi_enter();
+	else
+		irq_enter();
+
+	for (i = 0; i < cpuhw->n_counters; ++i) {
+		counter = cpuhw->counter[i];
+		if (!counter->hw.idx || is_limited_pmc(counter->hw.idx))
+			continue;
+		val = read_pmc(counter->hw.idx);
+		if ((int)val < 0) {
+			/* counter has overflowed */
+			found = 1;
+			record_and_restart(counter, val, regs, nmi);
+		}
+	}
+
+	/*
+	 * In case we didn't find and reset the counter that caused
+	 * the interrupt, scan all counters and reset any that are
+	 * negative, to avoid getting continual interrupts.
+	 * Any that we processed in the previous loop will not be negative.
+	 */
+	if (!found) {
+		for (i = 0; i < ppmu->n_counter; ++i) {
+			if (is_limited_pmc(i + 1))
+				continue;
+			val = read_pmc(i + 1);
+			if ((int)val < 0)
+				write_pmc(i + 1, 0);
+		}
+	}
+
+	/*
+	 * Reset MMCR0 to its normal value.  This will set PMXE and
+	 * clear FC (freeze counters) and PMAO (perf mon alert occurred)
+	 * and thus allow interrupts to occur again.
+	 * XXX might want to use MSR.PM to keep the counters frozen until
+	 * we get back out of this interrupt.
+	 */
+	write_mmcr0(cpuhw, cpuhw->mmcr[0]);
+
+	if (nmi)
+		nmi_exit();
+	else
+		irq_exit();
+}
+
+void hw_perf_counter_setup(int cpu)
+{
+	struct cpu_hw_counters *cpuhw = &per_cpu(cpu_hw_counters, cpu);
+
+	memset(cpuhw, 0, sizeof(*cpuhw));
+	cpuhw->mmcr[0] = MMCR0_FC;
+}
+
+extern struct power_pmu power4_pmu;
+extern struct power_pmu ppc970_pmu;
+extern struct power_pmu power5_pmu;
+extern struct power_pmu power5p_pmu;
+extern struct power_pmu power6_pmu;
+extern struct power_pmu power7_pmu;
+
+static int init_perf_counters(void)
+{
+	unsigned long pvr;
+
+	/* XXX should get this from cputable */
+	pvr = mfspr(SPRN_PVR);
+	switch (PVR_VER(pvr)) {
+	case PV_POWER4:
+	case PV_POWER4p:
+		ppmu = &power4_pmu;
+		break;
+	case PV_970:
+	case PV_970FX:
+	case PV_970MP:
+		ppmu = &ppc970_pmu;
+		break;
+	case PV_POWER5:
+		ppmu = &power5_pmu;
+		break;
+	case PV_POWER5p:
+		ppmu = &power5p_pmu;
+		break;
+	case 0x3e:
+		ppmu = &power6_pmu;
+		break;
+	case 0x3f:
+		ppmu = &power7_pmu;
+		break;
+	}
+
+	/*
+	 * Use FCHV to ignore kernel events if MSR.HV is set.
+	 */
+	if (mfmsr() & MSR_HV)
+		freeze_counters_kernel = MMCR0_FCHV;
+
+	return 0;
+}
+
+arch_initcall(init_perf_counters);
diff --git a/arch/powerpc/kernel/power4-pmu.c b/arch/powerpc/kernel/power4-pmu.c
new file mode 100644
index 00000000000..07bd308a5fa
--- /dev/null
+++ b/arch/powerpc/kernel/power4-pmu.c
@@ -0,0 +1,598 @@
+/*
+ * Performance counter support for POWER4 (GP) and POWER4+ (GQ) processors.
+ *
+ * Copyright 2009 Paul Mackerras, IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+#include <linux/kernel.h>
+#include <linux/perf_counter.h>
+#include <asm/reg.h>
+
+/*
+ * Bits in event code for POWER4
+ */
+#define PM_PMC_SH	12	/* PMC number (1-based) for direct events */
+#define PM_PMC_MSK	0xf
+#define PM_UNIT_SH	8	/* TTMMUX number and setting - unit select */
+#define PM_UNIT_MSK	0xf
+#define PM_LOWER_SH	6
+#define PM_LOWER_MSK	1
+#define PM_LOWER_MSKS	0x40
+#define PM_BYTE_SH	4	/* Byte number of event bus to use */
+#define PM_BYTE_MSK	3
+#define PM_PMCSEL_MSK	7
+
+/*
+ * Unit code values
+ */
+#define PM_FPU		1
+#define PM_ISU1		2
+#define PM_IFU		3
+#define PM_IDU0		4
+#define PM_ISU1_ALT	6
+#define PM_ISU2		7
+#define PM_IFU_ALT	8
+#define PM_LSU0		9
+#define PM_LSU1		0xc
+#define PM_GPS		0xf
+
+/*
+ * Bits in MMCR0 for POWER4
+ */
+#define MMCR0_PMC1SEL_SH	8
+#define MMCR0_PMC2SEL_SH	1
+#define MMCR_PMCSEL_MSK		0x1f
+
+/*
+ * Bits in MMCR1 for POWER4
+ */
+#define MMCR1_TTM0SEL_SH	62
+#define MMCR1_TTC0SEL_SH	61
+#define MMCR1_TTM1SEL_SH	59
+#define MMCR1_TTC1SEL_SH	58
+#define MMCR1_TTM2SEL_SH	56
+#define MMCR1_TTC2SEL_SH	55
+#define MMCR1_TTM3SEL_SH	53
+#define MMCR1_TTC3SEL_SH	52
+#define MMCR1_TTMSEL_MSK	3
+#define MMCR1_TD_CP_DBG0SEL_SH	50
+#define MMCR1_TD_CP_DBG1SEL_SH	48
+#define MMCR1_TD_CP_DBG2SEL_SH	46
+#define MMCR1_TD_CP_DBG3SEL_SH	44
+#define MMCR1_DEBUG0SEL_SH	43
+#define MMCR1_DEBUG1SEL_SH	42
+#define MMCR1_DEBUG2SEL_SH	41
+#define MMCR1_DEBUG3SEL_SH	40
+#define MMCR1_PMC1_ADDER_SEL_SH	39
+#define MMCR1_PMC2_ADDER_SEL_SH	38
+#define MMCR1_PMC6_ADDER_SEL_SH	37
+#define MMCR1_PMC5_ADDER_SEL_SH	36
+#define MMCR1_PMC8_ADDER_SEL_SH	35
+#define MMCR1_PMC7_ADDER_SEL_SH	34
+#define MMCR1_PMC3_ADDER_SEL_SH	33
+#define MMCR1_PMC4_ADDER_SEL_SH	32
+#define MMCR1_PMC3SEL_SH	27
+#define MMCR1_PMC4SEL_SH	22
+#define MMCR1_PMC5SEL_SH	17
+#define MMCR1_PMC6SEL_SH	12
+#define MMCR1_PMC7SEL_SH	7
+#define MMCR1_PMC8SEL_SH	2	/* note bit 0 is in MMCRA for GP */
+
+static short mmcr1_adder_bits[8] = {
+	MMCR1_PMC1_ADDER_SEL_SH,
+	MMCR1_PMC2_ADDER_SEL_SH,
+	MMCR1_PMC3_ADDER_SEL_SH,
+	MMCR1_PMC4_ADDER_SEL_SH,
+	MMCR1_PMC5_ADDER_SEL_SH,
+	MMCR1_PMC6_ADDER_SEL_SH,
+	MMCR1_PMC7_ADDER_SEL_SH,
+	MMCR1_PMC8_ADDER_SEL_SH
+};
+
+/*
+ * Bits in MMCRA
+ */
+#define MMCRA_PMC8SEL0_SH	17	/* PMC8SEL bit 0 for GP */
+
+/*
+ * Layout of constraint bits:
+ * 6666555555555544444444443333333333222222222211111111110000000000
+ * 3210987654321098765432109876543210987654321098765432109876543210
+ *        |[  >[  >[   >|||[  >[  ><  ><  ><  ><  ><><><><><><><><>
+ *        | UC1 UC2 UC3 ||| PS1 PS2 B0  B1  B2  B3 P1P2P3P4P5P6P7P8
+ * 	  \SMPL	        ||\TTC3SEL
+ * 		        |\TTC_IFU_SEL
+ * 		        \TTM2SEL0
+ *
+ * SMPL - SAMPLE_ENABLE constraint
+ *     56: SAMPLE_ENABLE value 0x0100_0000_0000_0000
+ *
+ * UC1 - unit constraint 1: can't have all three of FPU/ISU1/IDU0|ISU2
+ *     55: UC1 error 0x0080_0000_0000_0000
+ *     54: FPU events needed 0x0040_0000_0000_0000
+ *     53: ISU1 events needed 0x0020_0000_0000_0000
+ *     52: IDU0|ISU2 events needed 0x0010_0000_0000_0000
+ *
+ * UC2 - unit constraint 2: can't have all three of FPU/IFU/LSU0
+ *     51: UC2 error 0x0008_0000_0000_0000
+ *     50: FPU events needed 0x0004_0000_0000_0000
+ *     49: IFU events needed 0x0002_0000_0000_0000
+ *     48: LSU0 events needed 0x0001_0000_0000_0000
+ *
+ * UC3 - unit constraint 3: can't have all four of LSU0/IFU/IDU0|ISU2/ISU1
+ *     47: UC3 error 0x8000_0000_0000
+ *     46: LSU0 events needed 0x4000_0000_0000
+ *     45: IFU events needed 0x2000_0000_0000
+ *     44: IDU0|ISU2 events needed 0x1000_0000_0000
+ *     43: ISU1 events needed 0x0800_0000_0000
+ *
+ * TTM2SEL0
+ *     42: 0 = IDU0 events needed
+ *     	   1 = ISU2 events needed 0x0400_0000_0000
+ *
+ * TTC_IFU_SEL
+ *     41: 0 = IFU.U events needed
+ *     	   1 = IFU.L events needed 0x0200_0000_0000
+ *
+ * TTC3SEL
+ *     40: 0 = LSU1.U events needed
+ *     	   1 = LSU1.L events needed 0x0100_0000_0000
+ *
+ * PS1
+ *     39: PS1 error 0x0080_0000_0000
+ *     36-38: count of events needing PMC1/2/5/6 0x0070_0000_0000
+ *
+ * PS2
+ *     35: PS2 error 0x0008_0000_0000
+ *     32-34: count of events needing PMC3/4/7/8 0x0007_0000_0000
+ *
+ * B0
+ *     28-31: Byte 0 event source 0xf000_0000
+ *     	   1 = FPU
+ * 	   2 = ISU1
+ * 	   3 = IFU
+ * 	   4 = IDU0
+ * 	   7 = ISU2
+ * 	   9 = LSU0
+ * 	   c = LSU1
+ * 	   f = GPS
+ *
+ * B1, B2, B3
+ *     24-27, 20-23, 16-19: Byte 1, 2, 3 event sources
+ *
+ * P8
+ *     15: P8 error 0x8000
+ *     14-15: Count of events needing PMC8
+ *
+ * P1..P7
+ *     0-13: Count of events needing PMC1..PMC7
+ *
+ * Note: this doesn't allow events using IFU.U to be combined with events
+ * using IFU.L, though that is feasible (using TTM0 and TTM2).  However
+ * there are no listed events for IFU.L (they are debug events not
+ * verified for performance monitoring) so this shouldn't cause a
+ * problem.
+ */
+
+static struct unitinfo {
+	u64	value, mask;
+	int	unit;
+	int	lowerbit;
+} p4_unitinfo[16] = {
+	[PM_FPU]  = { 0x44000000000000ull, 0x88000000000000ull, PM_FPU, 0 },
+	[PM_ISU1] = { 0x20080000000000ull, 0x88000000000000ull, PM_ISU1, 0 },
+	[PM_ISU1_ALT] =
+		    { 0x20080000000000ull, 0x88000000000000ull, PM_ISU1, 0 },
+	[PM_IFU]  = { 0x02200000000000ull, 0x08820000000000ull, PM_IFU, 41 },
+	[PM_IFU_ALT] =
+		    { 0x02200000000000ull, 0x08820000000000ull, PM_IFU, 41 },
+	[PM_IDU0] = { 0x10100000000000ull, 0x80840000000000ull, PM_IDU0, 1 },
+	[PM_ISU2] = { 0x10140000000000ull, 0x80840000000000ull, PM_ISU2, 0 },
+	[PM_LSU0] = { 0x01400000000000ull, 0x08800000000000ull, PM_LSU0, 0 },
+	[PM_LSU1] = { 0x00000000000000ull, 0x00010000000000ull, PM_LSU1, 40 },
+	[PM_GPS]  = { 0x00000000000000ull, 0x00000000000000ull, PM_GPS, 0 }
+};
+
+static unsigned char direct_marked_event[8] = {
+	(1<<2) | (1<<3),	/* PMC1: PM_MRK_GRP_DISP, PM_MRK_ST_CMPL */
+	(1<<3) | (1<<5),	/* PMC2: PM_THRESH_TIMEO, PM_MRK_BRU_FIN */
+	(1<<3),			/* PMC3: PM_MRK_ST_CMPL_INT */
+	(1<<4) | (1<<5),	/* PMC4: PM_MRK_GRP_CMPL, PM_MRK_CRU_FIN */
+	(1<<4) | (1<<5),	/* PMC5: PM_MRK_GRP_TIMEO */
+	(1<<3) | (1<<4) | (1<<5),
+		/* PMC6: PM_MRK_ST_GPS, PM_MRK_FXU_FIN, PM_MRK_GRP_ISSUED */
+	(1<<4) | (1<<5),	/* PMC7: PM_MRK_FPU_FIN, PM_MRK_INST_FIN */
+	(1<<4),			/* PMC8: PM_MRK_LSU_FIN */
+};
+
+/*
+ * Returns 1 if event counts things relating to marked instructions
+ * and thus needs the MMCRA_SAMPLE_ENABLE bit set, or 0 if not.
+ */
+static int p4_marked_instr_event(u64 event)
+{
+	int pmc, psel, unit, byte, bit;
+	unsigned int mask;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	psel = event & PM_PMCSEL_MSK;
+	if (pmc) {
+		if (direct_marked_event[pmc - 1] & (1 << psel))
+			return 1;
+		if (psel == 0)		/* add events */
+			bit = (pmc <= 4)? pmc - 1: 8 - pmc;
+		else if (psel == 6)	/* decode events */
+			bit = 4;
+		else
+			return 0;
+	} else
+		bit = psel;
+
+	byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+	unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+	mask = 0;
+	switch (unit) {
+	case PM_LSU1:
+		if (event & PM_LOWER_MSKS)
+			mask = 1 << 28;		/* byte 7 bit 4 */
+		else
+			mask = 6 << 24;		/* byte 3 bits 1 and 2 */
+		break;
+	case PM_LSU0:
+		/* byte 3, bit 3; byte 2 bits 0,2,3,4,5; byte 1 */
+		mask = 0x083dff00;
+	}
+	return (mask >> (byte * 8 + bit)) & 1;
+}
+
+static int p4_get_constraint(u64 event, u64 *maskp, u64 *valp)
+{
+	int pmc, byte, unit, lower, sh;
+	u64 mask = 0, value = 0;
+	int grp = -1;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	if (pmc) {
+		if (pmc > 8)
+			return -1;
+		sh = (pmc - 1) * 2;
+		mask |= 2 << sh;
+		value |= 1 << sh;
+		grp = ((pmc - 1) >> 1) & 1;
+	}
+	unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+	byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+	if (unit) {
+		lower = (event >> PM_LOWER_SH) & PM_LOWER_MSK;
+
+		/*
+		 * Bus events on bytes 0 and 2 can be counted
+		 * on PMC1/2/5/6; bytes 1 and 3 on PMC3/4/7/8.
+		 */
+		if (!pmc)
+			grp = byte & 1;
+
+		if (!p4_unitinfo[unit].unit)
+			return -1;
+		mask  |= p4_unitinfo[unit].mask;
+		value |= p4_unitinfo[unit].value;
+		sh = p4_unitinfo[unit].lowerbit;
+		if (sh > 1)
+			value |= (u64)lower << sh;
+		else if (lower != sh)
+			return -1;
+		unit = p4_unitinfo[unit].unit;
+
+		/* Set byte lane select field */
+		mask  |= 0xfULL << (28 - 4 * byte);
+		value |= (u64)unit << (28 - 4 * byte);
+	}
+	if (grp == 0) {
+		/* increment PMC1/2/5/6 field */
+		mask  |= 0x8000000000ull;
+		value |= 0x1000000000ull;
+	} else {
+		/* increment PMC3/4/7/8 field */
+		mask  |= 0x800000000ull;
+		value |= 0x100000000ull;
+	}
+
+	/* Marked instruction events need sample_enable set */
+	if (p4_marked_instr_event(event)) {
+		mask  |= 1ull << 56;
+		value |= 1ull << 56;
+	}
+
+	/* PMCSEL=6 decode events on byte 2 need sample_enable clear */
+	if (pmc && (event & PM_PMCSEL_MSK) == 6 && byte == 2)
+		mask  |= 1ull << 56;
+
+	*maskp = mask;
+	*valp = value;
+	return 0;
+}
+
+static unsigned int ppc_inst_cmpl[] = {
+	0x1001, 0x4001, 0x6001, 0x7001, 0x8001
+};
+
+static int p4_get_alternatives(u64 event, unsigned int flags, u64 alt[])
+{
+	int i, j, na;
+
+	alt[0] = event;
+	na = 1;
+
+	/* 2 possibilities for PM_GRP_DISP_REJECT */
+	if (event == 0x8003 || event == 0x0224) {
+		alt[1] = event ^ (0x8003 ^ 0x0224);
+		return 2;
+	}
+
+	/* 2 possibilities for PM_ST_MISS_L1 */
+	if (event == 0x0c13 || event == 0x0c23) {
+		alt[1] = event ^ (0x0c13 ^ 0x0c23);
+		return 2;
+	}
+
+	/* several possibilities for PM_INST_CMPL */
+	for (i = 0; i < ARRAY_SIZE(ppc_inst_cmpl); ++i) {
+		if (event == ppc_inst_cmpl[i]) {
+			for (j = 0; j < ARRAY_SIZE(ppc_inst_cmpl); ++j)
+				if (j != i)
+					alt[na++] = ppc_inst_cmpl[j];
+			break;
+		}
+	}
+
+	return na;
+}
+
+static int p4_compute_mmcr(u64 event[], int n_ev,
+			   unsigned int hwc[], u64 mmcr[])
+{
+	u64 mmcr0 = 0, mmcr1 = 0, mmcra = 0;
+	unsigned int pmc, unit, byte, psel, lower;
+	unsigned int ttm, grp;
+	unsigned int pmc_inuse = 0;
+	unsigned int pmc_grp_use[2];
+	unsigned char busbyte[4];
+	unsigned char unituse[16];
+	unsigned int unitlower = 0;
+	int i;
+
+	if (n_ev > 8)
+		return -1;
+
+	/* First pass to count resource use */
+	pmc_grp_use[0] = pmc_grp_use[1] = 0;
+	memset(busbyte, 0, sizeof(busbyte));
+	memset(unituse, 0, sizeof(unituse));
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		if (pmc) {
+			if (pmc_inuse & (1 << (pmc - 1)))
+				return -1;
+			pmc_inuse |= 1 << (pmc - 1);
+			/* count 1/2/5/6 vs 3/4/7/8 use */
+			++pmc_grp_use[((pmc - 1) >> 1) & 1];
+		}
+		unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+		byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
+		lower = (event[i] >> PM_LOWER_SH) & PM_LOWER_MSK;
+		if (unit) {
+			if (!pmc)
+				++pmc_grp_use[byte & 1];
+			if (unit == 6 || unit == 8)
+				/* map alt ISU1/IFU codes: 6->2, 8->3 */
+				unit = (unit >> 1) - 1;
+			if (busbyte[byte] && busbyte[byte] != unit)
+				return -1;
+			busbyte[byte] = unit;
+			lower <<= unit;
+			if (unituse[unit] && lower != (unitlower & lower))
+				return -1;
+			unituse[unit] = 1;
+			unitlower |= lower;
+		}
+	}
+	if (pmc_grp_use[0] > 4 || pmc_grp_use[1] > 4)
+		return -1;
+
+	/*
+	 * Assign resources and set multiplexer selects.
+	 *
+	 * Units 1,2,3 are on TTM0, 4,6,7 on TTM1, 8,10 on TTM2.
+	 * Each TTMx can only select one unit, but since
+	 * units 2 and 6 are both ISU1, and 3 and 8 are both IFU,
+	 * we have some choices.
+	 */
+	if (unituse[2] & (unituse[1] | (unituse[3] & unituse[9]))) {
+		unituse[6] = 1;		/* Move 2 to 6 */
+		unituse[2] = 0;
+	}
+	if (unituse[3] & (unituse[1] | unituse[2])) {
+		unituse[8] = 1;		/* Move 3 to 8 */
+		unituse[3] = 0;
+		unitlower = (unitlower & ~8) | ((unitlower & 8) << 5);
+	}
+	/* Check only one unit per TTMx */
+	if (unituse[1] + unituse[2] + unituse[3] > 1 ||
+	    unituse[4] + unituse[6] + unituse[7] > 1 ||
+	    unituse[8] + unituse[9] > 1 ||
+	    (unituse[5] | unituse[10] | unituse[11] |
+	     unituse[13] | unituse[14]))
+		return -1;
+
+	/* Set TTMxSEL fields.  Note, units 1-3 => TTM0SEL codes 0-2 */
+	mmcr1 |= (u64)(unituse[3] * 2 + unituse[2]) << MMCR1_TTM0SEL_SH;
+	mmcr1 |= (u64)(unituse[7] * 3 + unituse[6] * 2) << MMCR1_TTM1SEL_SH;
+	mmcr1 |= (u64)unituse[9] << MMCR1_TTM2SEL_SH;
+
+	/* Set TTCxSEL fields. */
+	if (unitlower & 0xe)
+		mmcr1 |= 1ull << MMCR1_TTC0SEL_SH;
+	if (unitlower & 0xf0)
+		mmcr1 |= 1ull << MMCR1_TTC1SEL_SH;
+	if (unitlower & 0xf00)
+		mmcr1 |= 1ull << MMCR1_TTC2SEL_SH;
+	if (unitlower & 0x7000)
+		mmcr1 |= 1ull << MMCR1_TTC3SEL_SH;
+
+	/* Set byte lane select fields. */
+	for (byte = 0; byte < 4; ++byte) {
+		unit = busbyte[byte];
+		if (!unit)
+			continue;
+		if (unit == 0xf) {
+			/* special case for GPS */
+			mmcr1 |= 1ull << (MMCR1_DEBUG0SEL_SH - byte);
+		} else {
+			if (!unituse[unit])
+				ttm = unit - 1;		/* 2->1, 3->2 */
+			else
+				ttm = unit >> 2;
+			mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2*byte);
+		}
+	}
+
+	/* Second pass: assign PMCs, set PMCxSEL and PMCx_ADDER_SEL fields */
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+		byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
+		psel = event[i] & PM_PMCSEL_MSK;
+		if (!pmc) {
+			/* Bus event or 00xxx direct event (off or cycles) */
+			if (unit)
+				psel |= 0x10 | ((byte & 2) << 2);
+			for (pmc = 0; pmc < 8; ++pmc) {
+				if (pmc_inuse & (1 << pmc))
+					continue;
+				grp = (pmc >> 1) & 1;
+				if (unit) {
+					if (grp == (byte & 1))
+						break;
+				} else if (pmc_grp_use[grp] < 4) {
+					++pmc_grp_use[grp];
+					break;
+				}
+			}
+			pmc_inuse |= 1 << pmc;
+		} else {
+			/* Direct event */
+			--pmc;
+			if (psel == 0 && (byte & 2))
+				/* add events on higher-numbered bus */
+				mmcr1 |= 1ull << mmcr1_adder_bits[pmc];
+			else if (psel == 6 && byte == 3)
+				/* seem to need to set sample_enable here */
+				mmcra |= MMCRA_SAMPLE_ENABLE;
+			psel |= 8;
+		}
+		if (pmc <= 1)
+			mmcr0 |= psel << (MMCR0_PMC1SEL_SH - 7 * pmc);
+		else
+			mmcr1 |= psel << (MMCR1_PMC3SEL_SH - 5 * (pmc - 2));
+		if (pmc == 7)	/* PMC8 */
+			mmcra |= (psel & 1) << MMCRA_PMC8SEL0_SH;
+		hwc[i] = pmc;
+		if (p4_marked_instr_event(event[i]))
+			mmcra |= MMCRA_SAMPLE_ENABLE;
+	}
+
+	if (pmc_inuse & 1)
+		mmcr0 |= MMCR0_PMC1CE;
+	if (pmc_inuse & 0xfe)
+		mmcr0 |= MMCR0_PMCjCE;
+
+	mmcra |= 0x2000;	/* mark only one IOP per PPC instruction */
+
+	/* Return MMCRx values */
+	mmcr[0] = mmcr0;
+	mmcr[1] = mmcr1;
+	mmcr[2] = mmcra;
+	return 0;
+}
+
+static void p4_disable_pmc(unsigned int pmc, u64 mmcr[])
+{
+	/*
+	 * Setting the PMCxSEL field to 0 disables PMC x.
+	 * (Note that pmc is 0-based here, not 1-based.)
+	 */
+	if (pmc <= 1) {
+		mmcr[0] &= ~(0x1fUL << (MMCR0_PMC1SEL_SH - 7 * pmc));
+	} else {
+		mmcr[1] &= ~(0x1fUL << (MMCR1_PMC3SEL_SH - 5 * (pmc - 2)));
+		if (pmc == 7)
+			mmcr[2] &= ~(1UL << MMCRA_PMC8SEL0_SH);
+	}
+}
+
+static int p4_generic_events[] = {
+	[PERF_COUNT_HW_CPU_CYCLES]		= 7,
+	[PERF_COUNT_HW_INSTRUCTIONS]		= 0x1001,
+	[PERF_COUNT_HW_CACHE_REFERENCES]	= 0x8c10, /* PM_LD_REF_L1 */
+	[PERF_COUNT_HW_CACHE_MISSES]		= 0x3c10, /* PM_LD_MISS_L1 */
+	[PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x330,  /* PM_BR_ISSUED */
+	[PERF_COUNT_HW_BRANCH_MISSES]		= 0x331,  /* PM_BR_MPRED_CR */
+};
+
+#define C(x)	PERF_COUNT_HW_CACHE_##x
+
+/*
+ * Table of generalized cache-related events.
+ * 0 means not supported, -1 means nonsensical, other values
+ * are event codes.
+ */
+static int power4_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
+	[C(L1D)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x8c10,		0x3c10	},
+		[C(OP_WRITE)] = {	0x7c10,		0xc13	},
+		[C(OP_PREFETCH)] = {	0xc35,		0	},
+	},
+	[C(L1I)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	0,		0	},
+	},
+	[C(LL)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0	},
+		[C(OP_WRITE)] = {	0,		0	},
+		[C(OP_PREFETCH)] = {	0xc34,		0	},
+	},
+	[C(DTLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x904	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+	[C(ITLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x900	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+	[C(BPU)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x330,		0x331	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+};
+
+struct power_pmu power4_pmu = {
+	.n_counter = 8,
+	.max_alternatives = 5,
+	.add_fields = 0x0000001100005555ull,
+	.test_adder = 0x0011083300000000ull,
+	.compute_mmcr = p4_compute_mmcr,
+	.get_constraint = p4_get_constraint,
+	.get_alternatives = p4_get_alternatives,
+	.disable_pmc = p4_disable_pmc,
+	.n_generic = ARRAY_SIZE(p4_generic_events),
+	.generic_events = p4_generic_events,
+	.cache_events = &power4_cache_events,
+};
diff --git a/arch/powerpc/kernel/power5+-pmu.c b/arch/powerpc/kernel/power5+-pmu.c
new file mode 100644
index 00000000000..41e5d2d958d
--- /dev/null
+++ b/arch/powerpc/kernel/power5+-pmu.c
@@ -0,0 +1,671 @@
+/*
+ * Performance counter support for POWER5+/++ (not POWER5) processors.
+ *
+ * Copyright 2009 Paul Mackerras, IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+#include <linux/kernel.h>
+#include <linux/perf_counter.h>
+#include <asm/reg.h>
+
+/*
+ * Bits in event code for POWER5+ (POWER5 GS) and POWER5++ (POWER5 GS DD3)
+ */
+#define PM_PMC_SH	20	/* PMC number (1-based) for direct events */
+#define PM_PMC_MSK	0xf
+#define PM_PMC_MSKS	(PM_PMC_MSK << PM_PMC_SH)
+#define PM_UNIT_SH	16	/* TTMMUX number and setting - unit select */
+#define PM_UNIT_MSK	0xf
+#define PM_BYTE_SH	12	/* Byte number of event bus to use */
+#define PM_BYTE_MSK	7
+#define PM_GRS_SH	8	/* Storage subsystem mux select */
+#define PM_GRS_MSK	7
+#define PM_BUSEVENT_MSK	0x80	/* Set if event uses event bus */
+#define PM_PMCSEL_MSK	0x7f
+
+/* Values in PM_UNIT field */
+#define PM_FPU		0
+#define PM_ISU0		1
+#define PM_IFU		2
+#define PM_ISU1		3
+#define PM_IDU		4
+#define PM_ISU0_ALT	6
+#define PM_GRS		7
+#define PM_LSU0		8
+#define PM_LSU1		0xc
+#define PM_LASTUNIT	0xc
+
+/*
+ * Bits in MMCR1 for POWER5+
+ */
+#define MMCR1_TTM0SEL_SH	62
+#define MMCR1_TTM1SEL_SH	60
+#define MMCR1_TTM2SEL_SH	58
+#define MMCR1_TTM3SEL_SH	56
+#define MMCR1_TTMSEL_MSK	3
+#define MMCR1_TD_CP_DBG0SEL_SH	54
+#define MMCR1_TD_CP_DBG1SEL_SH	52
+#define MMCR1_TD_CP_DBG2SEL_SH	50
+#define MMCR1_TD_CP_DBG3SEL_SH	48
+#define MMCR1_GRS_L2SEL_SH	46
+#define MMCR1_GRS_L2SEL_MSK	3
+#define MMCR1_GRS_L3SEL_SH	44
+#define MMCR1_GRS_L3SEL_MSK	3
+#define MMCR1_GRS_MCSEL_SH	41
+#define MMCR1_GRS_MCSEL_MSK	7
+#define MMCR1_GRS_FABSEL_SH	39
+#define MMCR1_GRS_FABSEL_MSK	3
+#define MMCR1_PMC1_ADDER_SEL_SH	35
+#define MMCR1_PMC2_ADDER_SEL_SH	34
+#define MMCR1_PMC3_ADDER_SEL_SH	33
+#define MMCR1_PMC4_ADDER_SEL_SH	32
+#define MMCR1_PMC1SEL_SH	25
+#define MMCR1_PMC2SEL_SH	17
+#define MMCR1_PMC3SEL_SH	9
+#define MMCR1_PMC4SEL_SH	1
+#define MMCR1_PMCSEL_SH(n)	(MMCR1_PMC1SEL_SH - (n) * 8)
+#define MMCR1_PMCSEL_MSK	0x7f
+
+/*
+ * Bits in MMCRA
+ */
+
+/*
+ * Layout of constraint bits:
+ * 6666555555555544444444443333333333222222222211111111110000000000
+ * 3210987654321098765432109876543210987654321098765432109876543210
+ *             [  ><><>< ><> <><>[  >  <  ><  ><  ><  ><><><><><><>
+ *             NC  G0G1G2 G3 T0T1 UC    B0  B1  B2  B3 P6P5P4P3P2P1
+ *
+ * NC - number of counters
+ *     51: NC error 0x0008_0000_0000_0000
+ *     48-50: number of events needing PMC1-4 0x0007_0000_0000_0000
+ *
+ * G0..G3 - GRS mux constraints
+ *     46-47: GRS_L2SEL value
+ *     44-45: GRS_L3SEL value
+ *     41-44: GRS_MCSEL value
+ *     39-40: GRS_FABSEL value
+ *	Note that these match up with their bit positions in MMCR1
+ *
+ * T0 - TTM0 constraint
+ *     36-37: TTM0SEL value (0=FPU, 2=IFU, 3=ISU1) 0x30_0000_0000
+ *
+ * T1 - TTM1 constraint
+ *     34-35: TTM1SEL value (0=IDU, 3=GRS) 0x0c_0000_0000
+ *
+ * UC - unit constraint: can't have all three of FPU|IFU|ISU1, ISU0, IDU|GRS
+ *     33: UC3 error 0x02_0000_0000
+ *     32: FPU|IFU|ISU1 events needed 0x01_0000_0000
+ *     31: ISU0 events needed 0x01_8000_0000
+ *     30: IDU|GRS events needed 0x00_4000_0000
+ *
+ * B0
+ *     24-27: Byte 0 event source 0x0f00_0000
+ *	      Encoding as for the event code
+ *
+ * B1, B2, B3
+ *     20-23, 16-19, 12-15: Byte 1, 2, 3 event sources
+ *
+ * P6
+ *     11: P6 error 0x800
+ *     10-11: Count of events needing PMC6
+ *
+ * P1..P5
+ *     0-9: Count of events needing PMC1..PMC5
+ */
+
+static const int grsel_shift[8] = {
+	MMCR1_GRS_L2SEL_SH, MMCR1_GRS_L2SEL_SH, MMCR1_GRS_L2SEL_SH,
+	MMCR1_GRS_L3SEL_SH, MMCR1_GRS_L3SEL_SH, MMCR1_GRS_L3SEL_SH,
+	MMCR1_GRS_MCSEL_SH, MMCR1_GRS_FABSEL_SH
+};
+
+/* Masks and values for using events from the various units */
+static u64 unit_cons[PM_LASTUNIT+1][2] = {
+	[PM_FPU] =   { 0x3200000000ull, 0x0100000000ull },
+	[PM_ISU0] =  { 0x0200000000ull, 0x0080000000ull },
+	[PM_ISU1] =  { 0x3200000000ull, 0x3100000000ull },
+	[PM_IFU] =   { 0x3200000000ull, 0x2100000000ull },
+	[PM_IDU] =   { 0x0e00000000ull, 0x0040000000ull },
+	[PM_GRS] =   { 0x0e00000000ull, 0x0c40000000ull },
+};
+
+static int power5p_get_constraint(u64 event, u64 *maskp, u64 *valp)
+{
+	int pmc, byte, unit, sh;
+	int bit, fmask;
+	u64 mask = 0, value = 0;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	if (pmc) {
+		if (pmc > 6)
+			return -1;
+		sh = (pmc - 1) * 2;
+		mask |= 2 << sh;
+		value |= 1 << sh;
+		if (pmc >= 5 && !(event == 0x500009 || event == 0x600005))
+			return -1;
+	}
+	if (event & PM_BUSEVENT_MSK) {
+		unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+		if (unit > PM_LASTUNIT)
+			return -1;
+		if (unit == PM_ISU0_ALT)
+			unit = PM_ISU0;
+		mask |= unit_cons[unit][0];
+		value |= unit_cons[unit][1];
+		byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+		if (byte >= 4) {
+			if (unit != PM_LSU1)
+				return -1;
+			/* Map LSU1 low word (bytes 4-7) to unit LSU1+1 */
+			++unit;
+			byte &= 3;
+		}
+		if (unit == PM_GRS) {
+			bit = event & 7;
+			fmask = (bit == 6)? 7: 3;
+			sh = grsel_shift[bit];
+			mask |= (u64)fmask << sh;
+			value |= (u64)((event >> PM_GRS_SH) & fmask) << sh;
+		}
+		/* Set byte lane select field */
+		mask  |= 0xfULL << (24 - 4 * byte);
+		value |= (u64)unit << (24 - 4 * byte);
+	}
+	if (pmc < 5) {
+		/* need a counter from PMC1-4 set */
+		mask  |= 0x8000000000000ull;
+		value |= 0x1000000000000ull;
+	}
+	*maskp = mask;
+	*valp = value;
+	return 0;
+}
+
+static int power5p_limited_pmc_event(u64 event)
+{
+	int pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+
+	return pmc == 5 || pmc == 6;
+}
+
+#define MAX_ALT	3	/* at most 3 alternatives for any event */
+
+static const unsigned int event_alternatives[][MAX_ALT] = {
+	{ 0x100c0,  0x40001f },			/* PM_GCT_FULL_CYC */
+	{ 0x120e4,  0x400002 },			/* PM_GRP_DISP_REJECT */
+	{ 0x230e2,  0x323087 },			/* PM_BR_PRED_CR */
+	{ 0x230e3,  0x223087, 0x3230a0 },	/* PM_BR_PRED_TA */
+	{ 0x410c7,  0x441084 },			/* PM_THRD_L2MISS_BOTH_CYC */
+	{ 0x800c4,  0xc20e0 },			/* PM_DTLB_MISS */
+	{ 0xc50c6,  0xc60e0 },			/* PM_MRK_DTLB_MISS */
+	{ 0x100005, 0x600005 },			/* PM_RUN_CYC */
+	{ 0x100009, 0x200009 },			/* PM_INST_CMPL */
+	{ 0x200015, 0x300015 },			/* PM_LSU_LMQ_SRQ_EMPTY_CYC */
+	{ 0x300009, 0x400009 },			/* PM_INST_DISP */
+};
+
+/*
+ * Scan the alternatives table for a match and return the
+ * index into the alternatives table if found, else -1.
+ */
+static int find_alternative(unsigned int event)
+{
+	int i, j;
+
+	for (i = 0; i < ARRAY_SIZE(event_alternatives); ++i) {
+		if (event < event_alternatives[i][0])
+			break;
+		for (j = 0; j < MAX_ALT && event_alternatives[i][j]; ++j)
+			if (event == event_alternatives[i][j])
+				return i;
+	}
+	return -1;
+}
+
+static const unsigned char bytedecode_alternatives[4][4] = {
+	/* PMC 1 */	{ 0x21, 0x23, 0x25, 0x27 },
+	/* PMC 2 */	{ 0x07, 0x17, 0x0e, 0x1e },
+	/* PMC 3 */	{ 0x20, 0x22, 0x24, 0x26 },
+	/* PMC 4 */	{ 0x07, 0x17, 0x0e, 0x1e }
+};
+
+/*
+ * Some direct events for decodes of event bus byte 3 have alternative
+ * PMCSEL values on other counters.  This returns the alternative
+ * event code for those that do, or -1 otherwise.  This also handles
+ * alternative PCMSEL values for add events.
+ */
+static s64 find_alternative_bdecode(u64 event)
+{
+	int pmc, altpmc, pp, j;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	if (pmc == 0 || pmc > 4)
+		return -1;
+	altpmc = 5 - pmc;	/* 1 <-> 4, 2 <-> 3 */
+	pp = event & PM_PMCSEL_MSK;
+	for (j = 0; j < 4; ++j) {
+		if (bytedecode_alternatives[pmc - 1][j] == pp) {
+			return (event & ~(PM_PMC_MSKS | PM_PMCSEL_MSK)) |
+				(altpmc << PM_PMC_SH) |
+				bytedecode_alternatives[altpmc - 1][j];
+		}
+	}
+
+	/* new decode alternatives for power5+ */
+	if (pmc == 1 && (pp == 0x0d || pp == 0x0e))
+		return event + (2 << PM_PMC_SH) + (0x2e - 0x0d);
+	if (pmc == 3 && (pp == 0x2e || pp == 0x2f))
+		return event - (2 << PM_PMC_SH) - (0x2e - 0x0d);
+
+	/* alternative add event encodings */
+	if (pp == 0x10 || pp == 0x28)
+		return ((event ^ (0x10 ^ 0x28)) & ~PM_PMC_MSKS) |
+			(altpmc << PM_PMC_SH);
+
+	return -1;
+}
+
+static int power5p_get_alternatives(u64 event, unsigned int flags, u64 alt[])
+{
+	int i, j, nalt = 1;
+	int nlim;
+	s64 ae;
+
+	alt[0] = event;
+	nalt = 1;
+	nlim = power5p_limited_pmc_event(event);
+	i = find_alternative(event);
+	if (i >= 0) {
+		for (j = 0; j < MAX_ALT; ++j) {
+			ae = event_alternatives[i][j];
+			if (ae && ae != event)
+				alt[nalt++] = ae;
+			nlim += power5p_limited_pmc_event(ae);
+		}
+	} else {
+		ae = find_alternative_bdecode(event);
+		if (ae > 0)
+			alt[nalt++] = ae;
+	}
+
+	if (flags & PPMU_ONLY_COUNT_RUN) {
+		/*
+		 * We're only counting in RUN state,
+		 * so PM_CYC is equivalent to PM_RUN_CYC
+		 * and PM_INST_CMPL === PM_RUN_INST_CMPL.
+		 * This doesn't include alternatives that don't provide
+		 * any extra flexibility in assigning PMCs (e.g.
+		 * 0x100005 for PM_RUN_CYC vs. 0xf for PM_CYC).
+		 * Note that even with these additional alternatives
+		 * we never end up with more than 3 alternatives for any event.
+		 */
+		j = nalt;
+		for (i = 0; i < nalt; ++i) {
+			switch (alt[i]) {
+			case 0xf:	/* PM_CYC */
+				alt[j++] = 0x600005;	/* PM_RUN_CYC */
+				++nlim;
+				break;
+			case 0x600005:	/* PM_RUN_CYC */
+				alt[j++] = 0xf;
+				break;
+			case 0x100009:	/* PM_INST_CMPL */
+				alt[j++] = 0x500009;	/* PM_RUN_INST_CMPL */
+				++nlim;
+				break;
+			case 0x500009:	/* PM_RUN_INST_CMPL */
+				alt[j++] = 0x100009;	/* PM_INST_CMPL */
+				alt[j++] = 0x200009;
+				break;
+			}
+		}
+		nalt = j;
+	}
+
+	if (!(flags & PPMU_LIMITED_PMC_OK) && nlim) {
+		/* remove the limited PMC events */
+		j = 0;
+		for (i = 0; i < nalt; ++i) {
+			if (!power5p_limited_pmc_event(alt[i])) {
+				alt[j] = alt[i];
+				++j;
+			}
+		}
+		nalt = j;
+	} else if ((flags & PPMU_LIMITED_PMC_REQD) && nlim < nalt) {
+		/* remove all but the limited PMC events */
+		j = 0;
+		for (i = 0; i < nalt; ++i) {
+			if (power5p_limited_pmc_event(alt[i])) {
+				alt[j] = alt[i];
+				++j;
+			}
+		}
+		nalt = j;
+	}
+
+	return nalt;
+}
+
+/*
+ * Map of which direct events on which PMCs are marked instruction events.
+ * Indexed by PMCSEL value, bit i (LE) set if PMC i is a marked event.
+ * Bit 0 is set if it is marked for all PMCs.
+ * The 0x80 bit indicates a byte decode PMCSEL value.
+ */
+static unsigned char direct_event_is_marked[0x28] = {
+	0,	/* 00 */
+	0x1f,	/* 01 PM_IOPS_CMPL */
+	0x2,	/* 02 PM_MRK_GRP_DISP */
+	0xe,	/* 03 PM_MRK_ST_CMPL, PM_MRK_ST_GPS, PM_MRK_ST_CMPL_INT */
+	0,	/* 04 */
+	0x1c,	/* 05 PM_MRK_BRU_FIN, PM_MRK_INST_FIN, PM_MRK_CRU_FIN */
+	0x80,	/* 06 */
+	0x80,	/* 07 */
+	0, 0, 0,/* 08 - 0a */
+	0x18,	/* 0b PM_THRESH_TIMEO, PM_MRK_GRP_TIMEO */
+	0,	/* 0c */
+	0x80,	/* 0d */
+	0x80,	/* 0e */
+	0,	/* 0f */
+	0,	/* 10 */
+	0x14,	/* 11 PM_MRK_GRP_BR_REDIR, PM_MRK_GRP_IC_MISS */
+	0,	/* 12 */
+	0x10,	/* 13 PM_MRK_GRP_CMPL */
+	0x1f,	/* 14 PM_GRP_MRK, PM_MRK_{FXU,FPU,LSU}_FIN */
+	0x2,	/* 15 PM_MRK_GRP_ISSUED */
+	0x80,	/* 16 */
+	0x80,	/* 17 */
+	0, 0, 0, 0, 0,
+	0x80,	/* 1d */
+	0x80,	/* 1e */
+	0,	/* 1f */
+	0x80,	/* 20 */
+	0x80,	/* 21 */
+	0x80,	/* 22 */
+	0x80,	/* 23 */
+	0x80,	/* 24 */
+	0x80,	/* 25 */
+	0x80,	/* 26 */
+	0x80,	/* 27 */
+};
+
+/*
+ * Returns 1 if event counts things relating to marked instructions
+ * and thus needs the MMCRA_SAMPLE_ENABLE bit set, or 0 if not.
+ */
+static int power5p_marked_instr_event(u64 event)
+{
+	int pmc, psel;
+	int bit, byte, unit;
+	u32 mask;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	psel = event & PM_PMCSEL_MSK;
+	if (pmc >= 5)
+		return 0;
+
+	bit = -1;
+	if (psel < sizeof(direct_event_is_marked)) {
+		if (direct_event_is_marked[psel] & (1 << pmc))
+			return 1;
+		if (direct_event_is_marked[psel] & 0x80)
+			bit = 4;
+		else if (psel == 0x08)
+			bit = pmc - 1;
+		else if (psel == 0x10)
+			bit = 4 - pmc;
+		else if (psel == 0x1b && (pmc == 1 || pmc == 3))
+			bit = 4;
+	} else if ((psel & 0x48) == 0x40) {
+		bit = psel & 7;
+	} else if (psel == 0x28) {
+		bit = pmc - 1;
+	} else if (pmc == 3 && (psel == 0x2e || psel == 0x2f)) {
+		bit = 4;
+	}
+
+	if (!(event & PM_BUSEVENT_MSK) || bit == -1)
+		return 0;
+
+	byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+	unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+	if (unit == PM_LSU0) {
+		/* byte 1 bits 0-7, byte 2 bits 0,2-4,6 */
+		mask = 0x5dff00;
+	} else if (unit == PM_LSU1 && byte >= 4) {
+		byte -= 4;
+		/* byte 5 bits 6-7, byte 6 bits 0,4, byte 7 bits 0-4,6 */
+		mask = 0x5f11c000;
+	} else
+		return 0;
+
+	return (mask >> (byte * 8 + bit)) & 1;
+}
+
+static int power5p_compute_mmcr(u64 event[], int n_ev,
+				unsigned int hwc[], u64 mmcr[])
+{
+	u64 mmcr1 = 0;
+	u64 mmcra = 0;
+	unsigned int pmc, unit, byte, psel;
+	unsigned int ttm;
+	int i, isbus, bit, grsel;
+	unsigned int pmc_inuse = 0;
+	unsigned char busbyte[4];
+	unsigned char unituse[16];
+	int ttmuse;
+
+	if (n_ev > 6)
+		return -1;
+
+	/* First pass to count resource use */
+	memset(busbyte, 0, sizeof(busbyte));
+	memset(unituse, 0, sizeof(unituse));
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		if (pmc) {
+			if (pmc > 6)
+				return -1;
+			if (pmc_inuse & (1 << (pmc - 1)))
+				return -1;
+			pmc_inuse |= 1 << (pmc - 1);
+		}
+		if (event[i] & PM_BUSEVENT_MSK) {
+			unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+			byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
+			if (unit > PM_LASTUNIT)
+				return -1;
+			if (unit == PM_ISU0_ALT)
+				unit = PM_ISU0;
+			if (byte >= 4) {
+				if (unit != PM_LSU1)
+					return -1;
+				++unit;
+				byte &= 3;
+			}
+			if (busbyte[byte] && busbyte[byte] != unit)
+				return -1;
+			busbyte[byte] = unit;
+			unituse[unit] = 1;
+		}
+	}
+
+	/*
+	 * Assign resources and set multiplexer selects.
+	 *
+	 * PM_ISU0 can go either on TTM0 or TTM1, but that's the only
+	 * choice we have to deal with.
+	 */
+	if (unituse[PM_ISU0] &
+	    (unituse[PM_FPU] | unituse[PM_IFU] | unituse[PM_ISU1])) {
+		unituse[PM_ISU0_ALT] = 1;	/* move ISU to TTM1 */
+		unituse[PM_ISU0] = 0;
+	}
+	/* Set TTM[01]SEL fields. */
+	ttmuse = 0;
+	for (i = PM_FPU; i <= PM_ISU1; ++i) {
+		if (!unituse[i])
+			continue;
+		if (ttmuse++)
+			return -1;
+		mmcr1 |= (u64)i << MMCR1_TTM0SEL_SH;
+	}
+	ttmuse = 0;
+	for (; i <= PM_GRS; ++i) {
+		if (!unituse[i])
+			continue;
+		if (ttmuse++)
+			return -1;
+		mmcr1 |= (u64)(i & 3) << MMCR1_TTM1SEL_SH;
+	}
+	if (ttmuse > 1)
+		return -1;
+
+	/* Set byte lane select fields, TTM[23]SEL and GRS_*SEL. */
+	for (byte = 0; byte < 4; ++byte) {
+		unit = busbyte[byte];
+		if (!unit)
+			continue;
+		if (unit == PM_ISU0 && unituse[PM_ISU0_ALT]) {
+			/* get ISU0 through TTM1 rather than TTM0 */
+			unit = PM_ISU0_ALT;
+		} else if (unit == PM_LSU1 + 1) {
+			/* select lower word of LSU1 for this byte */
+			mmcr1 |= 1ull << (MMCR1_TTM3SEL_SH + 3 - byte);
+		}
+		ttm = unit >> 2;
+		mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte);
+	}
+
+	/* Second pass: assign PMCs, set PMCxSEL and PMCx_ADDER_SEL fields */
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+		byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
+		psel = event[i] & PM_PMCSEL_MSK;
+		isbus = event[i] & PM_BUSEVENT_MSK;
+		if (!pmc) {
+			/* Bus event or any-PMC direct event */
+			for (pmc = 0; pmc < 4; ++pmc) {
+				if (!(pmc_inuse & (1 << pmc)))
+					break;
+			}
+			if (pmc >= 4)
+				return -1;
+			pmc_inuse |= 1 << pmc;
+		} else if (pmc <= 4) {
+			/* Direct event */
+			--pmc;
+			if (isbus && (byte & 2) &&
+			    (psel == 8 || psel == 0x10 || psel == 0x28))
+				/* add events on higher-numbered bus */
+				mmcr1 |= 1ull << (MMCR1_PMC1_ADDER_SEL_SH - pmc);
+		} else {
+			/* Instructions or run cycles on PMC5/6 */
+			--pmc;
+		}
+		if (isbus && unit == PM_GRS) {
+			bit = psel & 7;
+			grsel = (event[i] >> PM_GRS_SH) & PM_GRS_MSK;
+			mmcr1 |= (u64)grsel << grsel_shift[bit];
+		}
+		if (power5p_marked_instr_event(event[i]))
+			mmcra |= MMCRA_SAMPLE_ENABLE;
+		if ((psel & 0x58) == 0x40 && (byte & 1) != ((pmc >> 1) & 1))
+			/* select alternate byte lane */
+			psel |= 0x10;
+		if (pmc <= 3)
+			mmcr1 |= psel << MMCR1_PMCSEL_SH(pmc);
+		hwc[i] = pmc;
+	}
+
+	/* Return MMCRx values */
+	mmcr[0] = 0;
+	if (pmc_inuse & 1)
+		mmcr[0] = MMCR0_PMC1CE;
+	if (pmc_inuse & 0x3e)
+		mmcr[0] |= MMCR0_PMCjCE;
+	mmcr[1] = mmcr1;
+	mmcr[2] = mmcra;
+	return 0;
+}
+
+static void power5p_disable_pmc(unsigned int pmc, u64 mmcr[])
+{
+	if (pmc <= 3)
+		mmcr[1] &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc));
+}
+
+static int power5p_generic_events[] = {
+	[PERF_COUNT_HW_CPU_CYCLES]		= 0xf,
+	[PERF_COUNT_HW_INSTRUCTIONS]		= 0x100009,
+	[PERF_COUNT_HW_CACHE_REFERENCES]	= 0x1c10a8, /* LD_REF_L1 */
+	[PERF_COUNT_HW_CACHE_MISSES]		= 0x3c1088, /* LD_MISS_L1 */
+	[PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x230e4,  /* BR_ISSUED */
+	[PERF_COUNT_HW_BRANCH_MISSES]		= 0x230e5,  /* BR_MPRED_CR */
+};
+
+#define C(x)	PERF_COUNT_HW_CACHE_##x
+
+/*
+ * Table of generalized cache-related events.
+ * 0 means not supported, -1 means nonsensical, other values
+ * are event codes.
+ */
+static int power5p_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
+	[C(L1D)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x1c10a8,	0x3c1088	},
+		[C(OP_WRITE)] = {	0x2c10a8,	0xc10c3		},
+		[C(OP_PREFETCH)] = {	0xc70e7,	-1		},
+	},
+	[C(L1I)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	0,		0		},
+	},
+	[C(LL)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0		},
+		[C(OP_WRITE)] = {	0,		0		},
+		[C(OP_PREFETCH)] = {	0xc50c3,	0		},
+	},
+	[C(DTLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0xc20e4,	0x800c4		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+	[C(ITLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x800c0		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+	[C(BPU)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x230e4,	0x230e5		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+};
+
+struct power_pmu power5p_pmu = {
+	.n_counter = 6,
+	.max_alternatives = MAX_ALT,
+	.add_fields = 0x7000000000055ull,
+	.test_adder = 0x3000040000000ull,
+	.compute_mmcr = power5p_compute_mmcr,
+	.get_constraint = power5p_get_constraint,
+	.get_alternatives = power5p_get_alternatives,
+	.disable_pmc = power5p_disable_pmc,
+	.limited_pmc_event = power5p_limited_pmc_event,
+	.flags = PPMU_LIMITED_PMC5_6,
+	.n_generic = ARRAY_SIZE(power5p_generic_events),
+	.generic_events = power5p_generic_events,
+	.cache_events = &power5p_cache_events,
+};
diff --git a/arch/powerpc/kernel/power5-pmu.c b/arch/powerpc/kernel/power5-pmu.c
new file mode 100644
index 00000000000..05600b66221
--- /dev/null
+++ b/arch/powerpc/kernel/power5-pmu.c
@@ -0,0 +1,611 @@
+/*
+ * Performance counter support for POWER5 (not POWER5++) processors.
+ *
+ * Copyright 2009 Paul Mackerras, IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+#include <linux/kernel.h>
+#include <linux/perf_counter.h>
+#include <asm/reg.h>
+
+/*
+ * Bits in event code for POWER5 (not POWER5++)
+ */
+#define PM_PMC_SH	20	/* PMC number (1-based) for direct events */
+#define PM_PMC_MSK	0xf
+#define PM_PMC_MSKS	(PM_PMC_MSK << PM_PMC_SH)
+#define PM_UNIT_SH	16	/* TTMMUX number and setting - unit select */
+#define PM_UNIT_MSK	0xf
+#define PM_BYTE_SH	12	/* Byte number of event bus to use */
+#define PM_BYTE_MSK	7
+#define PM_GRS_SH	8	/* Storage subsystem mux select */
+#define PM_GRS_MSK	7
+#define PM_BUSEVENT_MSK	0x80	/* Set if event uses event bus */
+#define PM_PMCSEL_MSK	0x7f
+
+/* Values in PM_UNIT field */
+#define PM_FPU		0
+#define PM_ISU0		1
+#define PM_IFU		2
+#define PM_ISU1		3
+#define PM_IDU		4
+#define PM_ISU0_ALT	6
+#define PM_GRS		7
+#define PM_LSU0		8
+#define PM_LSU1		0xc
+#define PM_LASTUNIT	0xc
+
+/*
+ * Bits in MMCR1 for POWER5
+ */
+#define MMCR1_TTM0SEL_SH	62
+#define MMCR1_TTM1SEL_SH	60
+#define MMCR1_TTM2SEL_SH	58
+#define MMCR1_TTM3SEL_SH	56
+#define MMCR1_TTMSEL_MSK	3
+#define MMCR1_TD_CP_DBG0SEL_SH	54
+#define MMCR1_TD_CP_DBG1SEL_SH	52
+#define MMCR1_TD_CP_DBG2SEL_SH	50
+#define MMCR1_TD_CP_DBG3SEL_SH	48
+#define MMCR1_GRS_L2SEL_SH	46
+#define MMCR1_GRS_L2SEL_MSK	3
+#define MMCR1_GRS_L3SEL_SH	44
+#define MMCR1_GRS_L3SEL_MSK	3
+#define MMCR1_GRS_MCSEL_SH	41
+#define MMCR1_GRS_MCSEL_MSK	7
+#define MMCR1_GRS_FABSEL_SH	39
+#define MMCR1_GRS_FABSEL_MSK	3
+#define MMCR1_PMC1_ADDER_SEL_SH	35
+#define MMCR1_PMC2_ADDER_SEL_SH	34
+#define MMCR1_PMC3_ADDER_SEL_SH	33
+#define MMCR1_PMC4_ADDER_SEL_SH	32
+#define MMCR1_PMC1SEL_SH	25
+#define MMCR1_PMC2SEL_SH	17
+#define MMCR1_PMC3SEL_SH	9
+#define MMCR1_PMC4SEL_SH	1
+#define MMCR1_PMCSEL_SH(n)	(MMCR1_PMC1SEL_SH - (n) * 8)
+#define MMCR1_PMCSEL_MSK	0x7f
+
+/*
+ * Bits in MMCRA
+ */
+
+/*
+ * Layout of constraint bits:
+ * 6666555555555544444444443333333333222222222211111111110000000000
+ * 3210987654321098765432109876543210987654321098765432109876543210
+ *         <><>[  ><><>< ><> [  >[ >[ ><  ><  ><  ><  ><><><><><><>
+ *         T0T1 NC G0G1G2 G3  UC PS1PS2 B0  B1  B2  B3 P6P5P4P3P2P1
+ *
+ * T0 - TTM0 constraint
+ *     54-55: TTM0SEL value (0=FPU, 2=IFU, 3=ISU1) 0xc0_0000_0000_0000
+ *
+ * T1 - TTM1 constraint
+ *     52-53: TTM1SEL value (0=IDU, 3=GRS) 0x30_0000_0000_0000
+ *
+ * NC - number of counters
+ *     51: NC error 0x0008_0000_0000_0000
+ *     48-50: number of events needing PMC1-4 0x0007_0000_0000_0000
+ *
+ * G0..G3 - GRS mux constraints
+ *     46-47: GRS_L2SEL value
+ *     44-45: GRS_L3SEL value
+ *     41-44: GRS_MCSEL value
+ *     39-40: GRS_FABSEL value
+ *	Note that these match up with their bit positions in MMCR1
+ *
+ * UC - unit constraint: can't have all three of FPU|IFU|ISU1, ISU0, IDU|GRS
+ *     37: UC3 error 0x20_0000_0000
+ *     36: FPU|IFU|ISU1 events needed 0x10_0000_0000
+ *     35: ISU0 events needed 0x08_0000_0000
+ *     34: IDU|GRS events needed 0x04_0000_0000
+ *
+ * PS1
+ *     33: PS1 error 0x2_0000_0000
+ *     31-32: count of events needing PMC1/2 0x1_8000_0000
+ *
+ * PS2
+ *     30: PS2 error 0x4000_0000
+ *     28-29: count of events needing PMC3/4 0x3000_0000
+ *
+ * B0
+ *     24-27: Byte 0 event source 0x0f00_0000
+ *	      Encoding as for the event code
+ *
+ * B1, B2, B3
+ *     20-23, 16-19, 12-15: Byte 1, 2, 3 event sources
+ *
+ * P1..P6
+ *     0-11: Count of events needing PMC1..PMC6
+ */
+
+static const int grsel_shift[8] = {
+	MMCR1_GRS_L2SEL_SH, MMCR1_GRS_L2SEL_SH, MMCR1_GRS_L2SEL_SH,
+	MMCR1_GRS_L3SEL_SH, MMCR1_GRS_L3SEL_SH, MMCR1_GRS_L3SEL_SH,
+	MMCR1_GRS_MCSEL_SH, MMCR1_GRS_FABSEL_SH
+};
+
+/* Masks and values for using events from the various units */
+static u64 unit_cons[PM_LASTUNIT+1][2] = {
+	[PM_FPU] =   { 0xc0002000000000ull, 0x00001000000000ull },
+	[PM_ISU0] =  { 0x00002000000000ull, 0x00000800000000ull },
+	[PM_ISU1] =  { 0xc0002000000000ull, 0xc0001000000000ull },
+	[PM_IFU] =   { 0xc0002000000000ull, 0x80001000000000ull },
+	[PM_IDU] =   { 0x30002000000000ull, 0x00000400000000ull },
+	[PM_GRS] =   { 0x30002000000000ull, 0x30000400000000ull },
+};
+
+static int power5_get_constraint(u64 event, u64 *maskp, u64 *valp)
+{
+	int pmc, byte, unit, sh;
+	int bit, fmask;
+	u64 mask = 0, value = 0;
+	int grp = -1;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	if (pmc) {
+		if (pmc > 6)
+			return -1;
+		sh = (pmc - 1) * 2;
+		mask |= 2 << sh;
+		value |= 1 << sh;
+		if (pmc <= 4)
+			grp = (pmc - 1) >> 1;
+		else if (event != 0x500009 && event != 0x600005)
+			return -1;
+	}
+	if (event & PM_BUSEVENT_MSK) {
+		unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+		if (unit > PM_LASTUNIT)
+			return -1;
+		if (unit == PM_ISU0_ALT)
+			unit = PM_ISU0;
+		mask |= unit_cons[unit][0];
+		value |= unit_cons[unit][1];
+		byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+		if (byte >= 4) {
+			if (unit != PM_LSU1)
+				return -1;
+			/* Map LSU1 low word (bytes 4-7) to unit LSU1+1 */
+			++unit;
+			byte &= 3;
+		}
+		if (unit == PM_GRS) {
+			bit = event & 7;
+			fmask = (bit == 6)? 7: 3;
+			sh = grsel_shift[bit];
+			mask |= (u64)fmask << sh;
+			value |= (u64)((event >> PM_GRS_SH) & fmask) << sh;
+		}
+		/*
+		 * Bus events on bytes 0 and 2 can be counted
+		 * on PMC1/2; bytes 1 and 3 on PMC3/4.
+		 */
+		if (!pmc)
+			grp = byte & 1;
+		/* Set byte lane select field */
+		mask  |= 0xfULL << (24 - 4 * byte);
+		value |= (u64)unit << (24 - 4 * byte);
+	}
+	if (grp == 0) {
+		/* increment PMC1/2 field */
+		mask  |= 0x200000000ull;
+		value |= 0x080000000ull;
+	} else if (grp == 1) {
+		/* increment PMC3/4 field */
+		mask  |= 0x40000000ull;
+		value |= 0x10000000ull;
+	}
+	if (pmc < 5) {
+		/* need a counter from PMC1-4 set */
+		mask  |= 0x8000000000000ull;
+		value |= 0x1000000000000ull;
+	}
+	*maskp = mask;
+	*valp = value;
+	return 0;
+}
+
+#define MAX_ALT	3	/* at most 3 alternatives for any event */
+
+static const unsigned int event_alternatives[][MAX_ALT] = {
+	{ 0x120e4,  0x400002 },			/* PM_GRP_DISP_REJECT */
+	{ 0x410c7,  0x441084 },			/* PM_THRD_L2MISS_BOTH_CYC */
+	{ 0x100005, 0x600005 },			/* PM_RUN_CYC */
+	{ 0x100009, 0x200009, 0x500009 },	/* PM_INST_CMPL */
+	{ 0x300009, 0x400009 },			/* PM_INST_DISP */
+};
+
+/*
+ * Scan the alternatives table for a match and return the
+ * index into the alternatives table if found, else -1.
+ */
+static int find_alternative(u64 event)
+{
+	int i, j;
+
+	for (i = 0; i < ARRAY_SIZE(event_alternatives); ++i) {
+		if (event < event_alternatives[i][0])
+			break;
+		for (j = 0; j < MAX_ALT && event_alternatives[i][j]; ++j)
+			if (event == event_alternatives[i][j])
+				return i;
+	}
+	return -1;
+}
+
+static const unsigned char bytedecode_alternatives[4][4] = {
+	/* PMC 1 */	{ 0x21, 0x23, 0x25, 0x27 },
+	/* PMC 2 */	{ 0x07, 0x17, 0x0e, 0x1e },
+	/* PMC 3 */	{ 0x20, 0x22, 0x24, 0x26 },
+	/* PMC 4 */	{ 0x07, 0x17, 0x0e, 0x1e }
+};
+
+/*
+ * Some direct events for decodes of event bus byte 3 have alternative
+ * PMCSEL values on other counters.  This returns the alternative
+ * event code for those that do, or -1 otherwise.
+ */
+static s64 find_alternative_bdecode(u64 event)
+{
+	int pmc, altpmc, pp, j;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	if (pmc == 0 || pmc > 4)
+		return -1;
+	altpmc = 5 - pmc;	/* 1 <-> 4, 2 <-> 3 */
+	pp = event & PM_PMCSEL_MSK;
+	for (j = 0; j < 4; ++j) {
+		if (bytedecode_alternatives[pmc - 1][j] == pp) {
+			return (event & ~(PM_PMC_MSKS | PM_PMCSEL_MSK)) |
+				(altpmc << PM_PMC_SH) |
+				bytedecode_alternatives[altpmc - 1][j];
+		}
+	}
+	return -1;
+}
+
+static int power5_get_alternatives(u64 event, unsigned int flags, u64 alt[])
+{
+	int i, j, nalt = 1;
+	s64 ae;
+
+	alt[0] = event;
+	nalt = 1;
+	i = find_alternative(event);
+	if (i >= 0) {
+		for (j = 0; j < MAX_ALT; ++j) {
+			ae = event_alternatives[i][j];
+			if (ae && ae != event)
+				alt[nalt++] = ae;
+		}
+	} else {
+		ae = find_alternative_bdecode(event);
+		if (ae > 0)
+			alt[nalt++] = ae;
+	}
+	return nalt;
+}
+
+/*
+ * Map of which direct events on which PMCs are marked instruction events.
+ * Indexed by PMCSEL value, bit i (LE) set if PMC i is a marked event.
+ * Bit 0 is set if it is marked for all PMCs.
+ * The 0x80 bit indicates a byte decode PMCSEL value.
+ */
+static unsigned char direct_event_is_marked[0x28] = {
+	0,	/* 00 */
+	0x1f,	/* 01 PM_IOPS_CMPL */
+	0x2,	/* 02 PM_MRK_GRP_DISP */
+	0xe,	/* 03 PM_MRK_ST_CMPL, PM_MRK_ST_GPS, PM_MRK_ST_CMPL_INT */
+	0,	/* 04 */
+	0x1c,	/* 05 PM_MRK_BRU_FIN, PM_MRK_INST_FIN, PM_MRK_CRU_FIN */
+	0x80,	/* 06 */
+	0x80,	/* 07 */
+	0, 0, 0,/* 08 - 0a */
+	0x18,	/* 0b PM_THRESH_TIMEO, PM_MRK_GRP_TIMEO */
+	0,	/* 0c */
+	0x80,	/* 0d */
+	0x80,	/* 0e */
+	0,	/* 0f */
+	0,	/* 10 */
+	0x14,	/* 11 PM_MRK_GRP_BR_REDIR, PM_MRK_GRP_IC_MISS */
+	0,	/* 12 */
+	0x10,	/* 13 PM_MRK_GRP_CMPL */
+	0x1f,	/* 14 PM_GRP_MRK, PM_MRK_{FXU,FPU,LSU}_FIN */
+	0x2,	/* 15 PM_MRK_GRP_ISSUED */
+	0x80,	/* 16 */
+	0x80,	/* 17 */
+	0, 0, 0, 0, 0,
+	0x80,	/* 1d */
+	0x80,	/* 1e */
+	0,	/* 1f */
+	0x80,	/* 20 */
+	0x80,	/* 21 */
+	0x80,	/* 22 */
+	0x80,	/* 23 */
+	0x80,	/* 24 */
+	0x80,	/* 25 */
+	0x80,	/* 26 */
+	0x80,	/* 27 */
+};
+
+/*
+ * Returns 1 if event counts things relating to marked instructions
+ * and thus needs the MMCRA_SAMPLE_ENABLE bit set, or 0 if not.
+ */
+static int power5_marked_instr_event(u64 event)
+{
+	int pmc, psel;
+	int bit, byte, unit;
+	u32 mask;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	psel = event & PM_PMCSEL_MSK;
+	if (pmc >= 5)
+		return 0;
+
+	bit = -1;
+	if (psel < sizeof(direct_event_is_marked)) {
+		if (direct_event_is_marked[psel] & (1 << pmc))
+			return 1;
+		if (direct_event_is_marked[psel] & 0x80)
+			bit = 4;
+		else if (psel == 0x08)
+			bit = pmc - 1;
+		else if (psel == 0x10)
+			bit = 4 - pmc;
+		else if (psel == 0x1b && (pmc == 1 || pmc == 3))
+			bit = 4;
+	} else if ((psel & 0x58) == 0x40)
+		bit = psel & 7;
+
+	if (!(event & PM_BUSEVENT_MSK))
+		return 0;
+
+	byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+	unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+	if (unit == PM_LSU0) {
+		/* byte 1 bits 0-7, byte 2 bits 0,2-4,6 */
+		mask = 0x5dff00;
+	} else if (unit == PM_LSU1 && byte >= 4) {
+		byte -= 4;
+		/* byte 4 bits 1,3,5,7, byte 5 bits 6-7, byte 7 bits 0-4,6 */
+		mask = 0x5f00c0aa;
+	} else
+		return 0;
+
+	return (mask >> (byte * 8 + bit)) & 1;
+}
+
+static int power5_compute_mmcr(u64 event[], int n_ev,
+			       unsigned int hwc[], u64 mmcr[])
+{
+	u64 mmcr1 = 0;
+	u64 mmcra = 0;
+	unsigned int pmc, unit, byte, psel;
+	unsigned int ttm, grp;
+	int i, isbus, bit, grsel;
+	unsigned int pmc_inuse = 0;
+	unsigned int pmc_grp_use[2];
+	unsigned char busbyte[4];
+	unsigned char unituse[16];
+	int ttmuse;
+
+	if (n_ev > 6)
+		return -1;
+
+	/* First pass to count resource use */
+	pmc_grp_use[0] = pmc_grp_use[1] = 0;
+	memset(busbyte, 0, sizeof(busbyte));
+	memset(unituse, 0, sizeof(unituse));
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		if (pmc) {
+			if (pmc > 6)
+				return -1;
+			if (pmc_inuse & (1 << (pmc - 1)))
+				return -1;
+			pmc_inuse |= 1 << (pmc - 1);
+			/* count 1/2 vs 3/4 use */
+			if (pmc <= 4)
+				++pmc_grp_use[(pmc - 1) >> 1];
+		}
+		if (event[i] & PM_BUSEVENT_MSK) {
+			unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+			byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
+			if (unit > PM_LASTUNIT)
+				return -1;
+			if (unit == PM_ISU0_ALT)
+				unit = PM_ISU0;
+			if (byte >= 4) {
+				if (unit != PM_LSU1)
+					return -1;
+				++unit;
+				byte &= 3;
+			}
+			if (!pmc)
+				++pmc_grp_use[byte & 1];
+			if (busbyte[byte] && busbyte[byte] != unit)
+				return -1;
+			busbyte[byte] = unit;
+			unituse[unit] = 1;
+		}
+	}
+	if (pmc_grp_use[0] > 2 || pmc_grp_use[1] > 2)
+		return -1;
+
+	/*
+	 * Assign resources and set multiplexer selects.
+	 *
+	 * PM_ISU0 can go either on TTM0 or TTM1, but that's the only
+	 * choice we have to deal with.
+	 */
+	if (unituse[PM_ISU0] &
+	    (unituse[PM_FPU] | unituse[PM_IFU] | unituse[PM_ISU1])) {
+		unituse[PM_ISU0_ALT] = 1;	/* move ISU to TTM1 */
+		unituse[PM_ISU0] = 0;
+	}
+	/* Set TTM[01]SEL fields. */
+	ttmuse = 0;
+	for (i = PM_FPU; i <= PM_ISU1; ++i) {
+		if (!unituse[i])
+			continue;
+		if (ttmuse++)
+			return -1;
+		mmcr1 |= (u64)i << MMCR1_TTM0SEL_SH;
+	}
+	ttmuse = 0;
+	for (; i <= PM_GRS; ++i) {
+		if (!unituse[i])
+			continue;
+		if (ttmuse++)
+			return -1;
+		mmcr1 |= (u64)(i & 3) << MMCR1_TTM1SEL_SH;
+	}
+	if (ttmuse > 1)
+		return -1;
+
+	/* Set byte lane select fields, TTM[23]SEL and GRS_*SEL. */
+	for (byte = 0; byte < 4; ++byte) {
+		unit = busbyte[byte];
+		if (!unit)
+			continue;
+		if (unit == PM_ISU0 && unituse[PM_ISU0_ALT]) {
+			/* get ISU0 through TTM1 rather than TTM0 */
+			unit = PM_ISU0_ALT;
+		} else if (unit == PM_LSU1 + 1) {
+			/* select lower word of LSU1 for this byte */
+			mmcr1 |= 1ull << (MMCR1_TTM3SEL_SH + 3 - byte);
+		}
+		ttm = unit >> 2;
+		mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte);
+	}
+
+	/* Second pass: assign PMCs, set PMCxSEL and PMCx_ADDER_SEL fields */
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+		byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
+		psel = event[i] & PM_PMCSEL_MSK;
+		isbus = event[i] & PM_BUSEVENT_MSK;
+		if (!pmc) {
+			/* Bus event or any-PMC direct event */
+			for (pmc = 0; pmc < 4; ++pmc) {
+				if (pmc_inuse & (1 << pmc))
+					continue;
+				grp = (pmc >> 1) & 1;
+				if (isbus) {
+					if (grp == (byte & 1))
+						break;
+				} else if (pmc_grp_use[grp] < 2) {
+					++pmc_grp_use[grp];
+					break;
+				}
+			}
+			pmc_inuse |= 1 << pmc;
+		} else if (pmc <= 4) {
+			/* Direct event */
+			--pmc;
+			if ((psel == 8 || psel == 0x10) && isbus && (byte & 2))
+				/* add events on higher-numbered bus */
+				mmcr1 |= 1ull << (MMCR1_PMC1_ADDER_SEL_SH - pmc);
+		} else {
+			/* Instructions or run cycles on PMC5/6 */
+			--pmc;
+		}
+		if (isbus && unit == PM_GRS) {
+			bit = psel & 7;
+			grsel = (event[i] >> PM_GRS_SH) & PM_GRS_MSK;
+			mmcr1 |= (u64)grsel << grsel_shift[bit];
+		}
+		if (power5_marked_instr_event(event[i]))
+			mmcra |= MMCRA_SAMPLE_ENABLE;
+		if (pmc <= 3)
+			mmcr1 |= psel << MMCR1_PMCSEL_SH(pmc);
+		hwc[i] = pmc;
+	}
+
+	/* Return MMCRx values */
+	mmcr[0] = 0;
+	if (pmc_inuse & 1)
+		mmcr[0] = MMCR0_PMC1CE;
+	if (pmc_inuse & 0x3e)
+		mmcr[0] |= MMCR0_PMCjCE;
+	mmcr[1] = mmcr1;
+	mmcr[2] = mmcra;
+	return 0;
+}
+
+static void power5_disable_pmc(unsigned int pmc, u64 mmcr[])
+{
+	if (pmc <= 3)
+		mmcr[1] &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc));
+}
+
+static int power5_generic_events[] = {
+	[PERF_COUNT_HW_CPU_CYCLES]		= 0xf,
+	[PERF_COUNT_HW_INSTRUCTIONS]		= 0x100009,
+	[PERF_COUNT_HW_CACHE_REFERENCES]	= 0x4c1090, /* LD_REF_L1 */
+	[PERF_COUNT_HW_CACHE_MISSES]		= 0x3c1088, /* LD_MISS_L1 */
+	[PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x230e4,  /* BR_ISSUED */
+	[PERF_COUNT_HW_BRANCH_MISSES]		= 0x230e5,  /* BR_MPRED_CR */
+};
+
+#define C(x)	PERF_COUNT_HW_CACHE_##x
+
+/*
+ * Table of generalized cache-related events.
+ * 0 means not supported, -1 means nonsensical, other values
+ * are event codes.
+ */
+static int power5_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
+	[C(L1D)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x4c1090,	0x3c1088	},
+		[C(OP_WRITE)] = {	0x3c1090,	0xc10c3		},
+		[C(OP_PREFETCH)] = {	0xc70e7,	0		},
+	},
+	[C(L1I)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	0,		0		},
+	},
+	[C(LL)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x3c309b	},
+		[C(OP_WRITE)] = {	0,		0		},
+		[C(OP_PREFETCH)] = {	0xc50c3,	0		},
+	},
+	[C(DTLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x2c4090,	0x800c4		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+	[C(ITLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x800c0		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+	[C(BPU)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x230e4,	0x230e5		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+};
+
+struct power_pmu power5_pmu = {
+	.n_counter = 6,
+	.max_alternatives = MAX_ALT,
+	.add_fields = 0x7000090000555ull,
+	.test_adder = 0x3000490000000ull,
+	.compute_mmcr = power5_compute_mmcr,
+	.get_constraint = power5_get_constraint,
+	.get_alternatives = power5_get_alternatives,
+	.disable_pmc = power5_disable_pmc,
+	.n_generic = ARRAY_SIZE(power5_generic_events),
+	.generic_events = power5_generic_events,
+	.cache_events = &power5_cache_events,
+};
diff --git a/arch/powerpc/kernel/power6-pmu.c b/arch/powerpc/kernel/power6-pmu.c
new file mode 100644
index 00000000000..46f74bebcfd
--- /dev/null
+++ b/arch/powerpc/kernel/power6-pmu.c
@@ -0,0 +1,532 @@
+/*
+ * Performance counter support for POWER6 processors.
+ *
+ * Copyright 2008-2009 Paul Mackerras, IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+#include <linux/kernel.h>
+#include <linux/perf_counter.h>
+#include <asm/reg.h>
+
+/*
+ * Bits in event code for POWER6
+ */
+#define PM_PMC_SH	20	/* PMC number (1-based) for direct events */
+#define PM_PMC_MSK	0x7
+#define PM_PMC_MSKS	(PM_PMC_MSK << PM_PMC_SH)
+#define PM_UNIT_SH	16	/* Unit event comes (TTMxSEL encoding) */
+#define PM_UNIT_MSK	0xf
+#define PM_UNIT_MSKS	(PM_UNIT_MSK << PM_UNIT_SH)
+#define PM_LLAV		0x8000	/* Load lookahead match value */
+#define PM_LLA		0x4000	/* Load lookahead match enable */
+#define PM_BYTE_SH	12	/* Byte of event bus to use */
+#define PM_BYTE_MSK	3
+#define PM_SUBUNIT_SH	8	/* Subunit event comes from (NEST_SEL enc.) */
+#define PM_SUBUNIT_MSK	7
+#define PM_SUBUNIT_MSKS	(PM_SUBUNIT_MSK << PM_SUBUNIT_SH)
+#define PM_PMCSEL_MSK	0xff	/* PMCxSEL value */
+#define PM_BUSEVENT_MSK	0xf3700
+
+/*
+ * Bits in MMCR1 for POWER6
+ */
+#define MMCR1_TTM0SEL_SH	60
+#define MMCR1_TTMSEL_SH(n)	(MMCR1_TTM0SEL_SH - (n) * 4)
+#define MMCR1_TTMSEL_MSK	0xf
+#define MMCR1_TTMSEL(m, n)	(((m) >> MMCR1_TTMSEL_SH(n)) & MMCR1_TTMSEL_MSK)
+#define MMCR1_NESTSEL_SH	45
+#define MMCR1_NESTSEL_MSK	0x7
+#define MMCR1_NESTSEL(m)	(((m) >> MMCR1_NESTSEL_SH) & MMCR1_NESTSEL_MSK)
+#define MMCR1_PMC1_LLA		((u64)1 << 44)
+#define MMCR1_PMC1_LLA_VALUE	((u64)1 << 39)
+#define MMCR1_PMC1_ADDR_SEL	((u64)1 << 35)
+#define MMCR1_PMC1SEL_SH	24
+#define MMCR1_PMCSEL_SH(n)	(MMCR1_PMC1SEL_SH - (n) * 8)
+#define MMCR1_PMCSEL_MSK	0xff
+
+/*
+ * Map of which direct events on which PMCs are marked instruction events.
+ * Indexed by PMCSEL value >> 1.
+ * Bottom 4 bits are a map of which PMCs are interesting,
+ * top 4 bits say what sort of event:
+ *   0 = direct marked event,
+ *   1 = byte decode event,
+ *   4 = add/and event (PMC1 -> bits 0 & 4),
+ *   5 = add/and event (PMC1 -> bits 1 & 5),
+ *   6 = add/and event (PMC1 -> bits 2 & 6),
+ *   7 = add/and event (PMC1 -> bits 3 & 7).
+ */
+static unsigned char direct_event_is_marked[0x60 >> 1] = {
+	0,	/* 00 */
+	0,	/* 02 */
+	0,	/* 04 */
+	0x07,	/* 06 PM_MRK_ST_CMPL, PM_MRK_ST_GPS, PM_MRK_ST_CMPL_INT */
+	0x04,	/* 08 PM_MRK_DFU_FIN */
+	0x06,	/* 0a PM_MRK_IFU_FIN, PM_MRK_INST_FIN */
+	0,	/* 0c */
+	0,	/* 0e */
+	0x02,	/* 10 PM_MRK_INST_DISP */
+	0x08,	/* 12 PM_MRK_LSU_DERAT_MISS */
+	0,	/* 14 */
+	0,	/* 16 */
+	0x0c,	/* 18 PM_THRESH_TIMEO, PM_MRK_INST_FIN */
+	0x0f,	/* 1a PM_MRK_INST_DISP, PM_MRK_{FXU,FPU,LSU}_FIN */
+	0x01,	/* 1c PM_MRK_INST_ISSUED */
+	0,	/* 1e */
+	0,	/* 20 */
+	0,	/* 22 */
+	0,	/* 24 */
+	0,	/* 26 */
+	0x15,	/* 28 PM_MRK_DATA_FROM_L2MISS, PM_MRK_DATA_FROM_L3MISS */
+	0,	/* 2a */
+	0,	/* 2c */
+	0,	/* 2e */
+	0x4f,	/* 30 */
+	0x7f,	/* 32 */
+	0x4f,	/* 34 */
+	0x5f,	/* 36 */
+	0x6f,	/* 38 */
+	0x4f,	/* 3a */
+	0,	/* 3c */
+	0x08,	/* 3e PM_MRK_INST_TIMEO */
+	0x1f,	/* 40 */
+	0x1f,	/* 42 */
+	0x1f,	/* 44 */
+	0x1f,	/* 46 */
+	0x1f,	/* 48 */
+	0x1f,	/* 4a */
+	0x1f,	/* 4c */
+	0x1f,	/* 4e */
+	0,	/* 50 */
+	0x05,	/* 52 PM_MRK_BR_TAKEN, PM_MRK_BR_MPRED */
+	0x1c,	/* 54 PM_MRK_PTEG_FROM_L3MISS, PM_MRK_PTEG_FROM_L2MISS */
+	0x02,	/* 56 PM_MRK_LD_MISS_L1 */
+	0,	/* 58 */
+	0,	/* 5a */
+	0,	/* 5c */
+	0,	/* 5e */
+};
+
+/*
+ * Masks showing for each unit which bits are marked events.
+ * These masks are in LE order, i.e. 0x00000001 is byte 0, bit 0.
+ */
+static u32 marked_bus_events[16] = {
+	0x01000000,	/* direct events set 1: byte 3 bit 0 */
+	0x00010000,	/* direct events set 2: byte 2 bit 0 */
+	0, 0, 0, 0,	/* IDU, IFU, nest: nothing */
+	0x00000088,	/* VMX set 1: byte 0 bits 3, 7 */
+	0x000000c0,	/* VMX set 2: byte 0 bits 4-7 */
+	0x04010000,	/* LSU set 1: byte 2 bit 0, byte 3 bit 2 */
+	0xff010000u,	/* LSU set 2: byte 2 bit 0, all of byte 3 */
+	0,		/* LSU set 3 */
+	0x00000010,	/* VMX set 3: byte 0 bit 4 */
+	0,		/* BFP set 1 */
+	0x00000022,	/* BFP set 2: byte 0 bits 1, 5 */
+	0, 0
+};
+	
+/*
+ * Returns 1 if event counts things relating to marked instructions
+ * and thus needs the MMCRA_SAMPLE_ENABLE bit set, or 0 if not.
+ */
+static int power6_marked_instr_event(u64 event)
+{
+	int pmc, psel, ptype;
+	int bit, byte, unit;
+	u32 mask;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	psel = (event & PM_PMCSEL_MSK) >> 1;	/* drop edge/level bit */
+	if (pmc >= 5)
+		return 0;
+
+	bit = -1;
+	if (psel < sizeof(direct_event_is_marked)) {
+		ptype = direct_event_is_marked[psel];
+		if (pmc == 0 || !(ptype & (1 << (pmc - 1))))
+			return 0;
+		ptype >>= 4;
+		if (ptype == 0)
+			return 1;
+		if (ptype == 1)
+			bit = 0;
+		else
+			bit = ptype ^ (pmc - 1);
+	} else if ((psel & 0x48) == 0x40)
+		bit = psel & 7;
+
+	if (!(event & PM_BUSEVENT_MSK) || bit == -1)
+		return 0;
+
+	byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+	unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+	mask = marked_bus_events[unit];
+	return (mask >> (byte * 8 + bit)) & 1;
+}
+
+/*
+ * Assign PMC numbers and compute MMCR1 value for a set of events
+ */
+static int p6_compute_mmcr(u64 event[], int n_ev,
+			   unsigned int hwc[], u64 mmcr[])
+{
+	u64 mmcr1 = 0;
+	u64 mmcra = 0;
+	int i;
+	unsigned int pmc, ev, b, u, s, psel;
+	unsigned int ttmset = 0;
+	unsigned int pmc_inuse = 0;
+
+	if (n_ev > 6)
+		return -1;
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		if (pmc) {
+			if (pmc_inuse & (1 << (pmc - 1)))
+				return -1;	/* collision! */
+			pmc_inuse |= 1 << (pmc - 1);
+		}
+	}
+	for (i = 0; i < n_ev; ++i) {
+		ev = event[i];
+		pmc = (ev >> PM_PMC_SH) & PM_PMC_MSK;
+		if (pmc) {
+			--pmc;
+		} else {
+			/* can go on any PMC; find a free one */
+			for (pmc = 0; pmc < 4; ++pmc)
+				if (!(pmc_inuse & (1 << pmc)))
+					break;
+			if (pmc >= 4)
+				return -1;
+			pmc_inuse |= 1 << pmc;
+		}
+		hwc[i] = pmc;
+		psel = ev & PM_PMCSEL_MSK;
+		if (ev & PM_BUSEVENT_MSK) {
+			/* this event uses the event bus */
+			b = (ev >> PM_BYTE_SH) & PM_BYTE_MSK;
+			u = (ev >> PM_UNIT_SH) & PM_UNIT_MSK;
+			/* check for conflict on this byte of event bus */
+			if ((ttmset & (1 << b)) && MMCR1_TTMSEL(mmcr1, b) != u)
+				return -1;
+			mmcr1 |= (u64)u << MMCR1_TTMSEL_SH(b);
+			ttmset |= 1 << b;
+			if (u == 5) {
+				/* Nest events have a further mux */
+				s = (ev >> PM_SUBUNIT_SH) & PM_SUBUNIT_MSK;
+				if ((ttmset & 0x10) &&
+				    MMCR1_NESTSEL(mmcr1) != s)
+					return -1;
+				ttmset |= 0x10;
+				mmcr1 |= (u64)s << MMCR1_NESTSEL_SH;
+			}
+			if (0x30 <= psel && psel <= 0x3d) {
+				/* these need the PMCx_ADDR_SEL bits */
+				if (b >= 2)
+					mmcr1 |= MMCR1_PMC1_ADDR_SEL >> pmc;
+			}
+			/* bus select values are different for PMC3/4 */
+			if (pmc >= 2 && (psel & 0x90) == 0x80)
+				psel ^= 0x20;
+		}
+		if (ev & PM_LLA) {
+			mmcr1 |= MMCR1_PMC1_LLA >> pmc;
+			if (ev & PM_LLAV)
+				mmcr1 |= MMCR1_PMC1_LLA_VALUE >> pmc;
+		}
+		if (power6_marked_instr_event(event[i]))
+			mmcra |= MMCRA_SAMPLE_ENABLE;
+		if (pmc < 4)
+			mmcr1 |= (u64)psel << MMCR1_PMCSEL_SH(pmc);
+	}
+	mmcr[0] = 0;
+	if (pmc_inuse & 1)
+		mmcr[0] = MMCR0_PMC1CE;
+	if (pmc_inuse & 0xe)
+		mmcr[0] |= MMCR0_PMCjCE;
+	mmcr[1] = mmcr1;
+	mmcr[2] = mmcra;
+	return 0;
+}
+
+/*
+ * Layout of constraint bits:
+ *
+ *	0-1	add field: number of uses of PMC1 (max 1)
+ *	2-3, 4-5, 6-7, 8-9, 10-11: ditto for PMC2, 3, 4, 5, 6
+ *	12-15	add field: number of uses of PMC1-4 (max 4)
+ *	16-19	select field: unit on byte 0 of event bus
+ *	20-23, 24-27, 28-31 ditto for bytes 1, 2, 3
+ *	32-34	select field: nest (subunit) event selector
+ */
+static int p6_get_constraint(u64 event, u64 *maskp, u64 *valp)
+{
+	int pmc, byte, sh, subunit;
+	u64 mask = 0, value = 0;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	if (pmc) {
+		if (pmc > 4 && !(event == 0x500009 || event == 0x600005))
+			return -1;
+		sh = (pmc - 1) * 2;
+		mask |= 2 << sh;
+		value |= 1 << sh;
+	}
+	if (event & PM_BUSEVENT_MSK) {
+		byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+		sh = byte * 4 + (16 - PM_UNIT_SH);
+		mask |= PM_UNIT_MSKS << sh;
+		value |= (u64)(event & PM_UNIT_MSKS) << sh;
+		if ((event & PM_UNIT_MSKS) == (5 << PM_UNIT_SH)) {
+			subunit = (event >> PM_SUBUNIT_SH) & PM_SUBUNIT_MSK;
+			mask  |= (u64)PM_SUBUNIT_MSK << 32;
+			value |= (u64)subunit << 32;
+		}
+	}
+	if (pmc <= 4) {
+		mask  |= 0x8000;	/* add field for count of PMC1-4 uses */
+		value |= 0x1000;
+	}
+	*maskp = mask;
+	*valp = value;
+	return 0;
+}
+
+static int p6_limited_pmc_event(u64 event)
+{
+	int pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+
+	return pmc == 5 || pmc == 6;
+}
+
+#define MAX_ALT	4	/* at most 4 alternatives for any event */
+
+static const unsigned int event_alternatives[][MAX_ALT] = {
+	{ 0x0130e8, 0x2000f6, 0x3000fc },	/* PM_PTEG_RELOAD_VALID */
+	{ 0x080080, 0x10000d, 0x30000c, 0x4000f0 }, /* PM_LD_MISS_L1 */
+	{ 0x080088, 0x200054, 0x3000f0 },	/* PM_ST_MISS_L1 */
+	{ 0x10000a, 0x2000f4, 0x600005 },	/* PM_RUN_CYC */
+	{ 0x10000b, 0x2000f5 },			/* PM_RUN_COUNT */
+	{ 0x10000e, 0x400010 },			/* PM_PURR */
+	{ 0x100010, 0x4000f8 },			/* PM_FLUSH */
+	{ 0x10001a, 0x200010 },			/* PM_MRK_INST_DISP */
+	{ 0x100026, 0x3000f8 },			/* PM_TB_BIT_TRANS */
+	{ 0x100054, 0x2000f0 },			/* PM_ST_FIN */
+	{ 0x100056, 0x2000fc },			/* PM_L1_ICACHE_MISS */
+	{ 0x1000f0, 0x40000a },			/* PM_INST_IMC_MATCH_CMPL */
+	{ 0x1000f8, 0x200008 },			/* PM_GCT_EMPTY_CYC */
+	{ 0x1000fc, 0x400006 },			/* PM_LSU_DERAT_MISS_CYC */
+	{ 0x20000e, 0x400007 },			/* PM_LSU_DERAT_MISS */
+	{ 0x200012, 0x300012 },			/* PM_INST_DISP */
+	{ 0x2000f2, 0x3000f2 },			/* PM_INST_DISP */
+	{ 0x2000f8, 0x300010 },			/* PM_EXT_INT */
+	{ 0x2000fe, 0x300056 },			/* PM_DATA_FROM_L2MISS */
+	{ 0x2d0030, 0x30001a },			/* PM_MRK_FPU_FIN */
+	{ 0x30000a, 0x400018 },			/* PM_MRK_INST_FIN */
+	{ 0x3000f6, 0x40000e },			/* PM_L1_DCACHE_RELOAD_VALID */
+	{ 0x3000fe, 0x400056 },			/* PM_DATA_FROM_L3MISS */
+};
+
+/*
+ * This could be made more efficient with a binary search on
+ * a presorted list, if necessary
+ */
+static int find_alternatives_list(u64 event)
+{
+	int i, j;
+	unsigned int alt;
+
+	for (i = 0; i < ARRAY_SIZE(event_alternatives); ++i) {
+		if (event < event_alternatives[i][0])
+			return -1;
+		for (j = 0; j < MAX_ALT; ++j) {
+			alt = event_alternatives[i][j];
+			if (!alt || event < alt)
+				break;
+			if (event == alt)
+				return i;
+		}
+	}
+	return -1;
+}
+
+static int p6_get_alternatives(u64 event, unsigned int flags, u64 alt[])
+{
+	int i, j, nlim;
+	unsigned int psel, pmc;
+	unsigned int nalt = 1;
+	u64 aevent;
+
+	alt[0] = event;
+	nlim = p6_limited_pmc_event(event);
+
+	/* check the alternatives table */
+	i = find_alternatives_list(event);
+	if (i >= 0) {
+		/* copy out alternatives from list */
+		for (j = 0; j < MAX_ALT; ++j) {
+			aevent = event_alternatives[i][j];
+			if (!aevent)
+				break;
+			if (aevent != event)
+				alt[nalt++] = aevent;
+			nlim += p6_limited_pmc_event(aevent);
+		}
+
+	} else {
+		/* Check for alternative ways of computing sum events */
+		/* PMCSEL 0x32 counter N == PMCSEL 0x34 counter 5-N */
+		psel = event & (PM_PMCSEL_MSK & ~1);	/* ignore edge bit */
+		pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+		if (pmc && (psel == 0x32 || psel == 0x34))
+			alt[nalt++] = ((event ^ 0x6) & ~PM_PMC_MSKS) |
+				((5 - pmc) << PM_PMC_SH);
+
+		/* PMCSEL 0x38 counter N == PMCSEL 0x3a counter N+/-2 */
+		if (pmc && (psel == 0x38 || psel == 0x3a))
+			alt[nalt++] = ((event ^ 0x2) & ~PM_PMC_MSKS) |
+				((pmc > 2? pmc - 2: pmc + 2) << PM_PMC_SH);
+	}
+
+	if (flags & PPMU_ONLY_COUNT_RUN) {
+		/*
+		 * We're only counting in RUN state,
+		 * so PM_CYC is equivalent to PM_RUN_CYC,
+		 * PM_INST_CMPL === PM_RUN_INST_CMPL, PM_PURR === PM_RUN_PURR.
+		 * This doesn't include alternatives that don't provide
+		 * any extra flexibility in assigning PMCs (e.g.
+		 * 0x10000a for PM_RUN_CYC vs. 0x1e for PM_CYC).
+		 * Note that even with these additional alternatives
+		 * we never end up with more than 4 alternatives for any event.
+		 */
+		j = nalt;
+		for (i = 0; i < nalt; ++i) {
+			switch (alt[i]) {
+			case 0x1e:	/* PM_CYC */
+				alt[j++] = 0x600005;	/* PM_RUN_CYC */
+				++nlim;
+				break;
+			case 0x10000a:	/* PM_RUN_CYC */
+				alt[j++] = 0x1e;	/* PM_CYC */
+				break;
+			case 2:		/* PM_INST_CMPL */
+				alt[j++] = 0x500009;	/* PM_RUN_INST_CMPL */
+				++nlim;
+				break;
+			case 0x500009:	/* PM_RUN_INST_CMPL */
+				alt[j++] = 2;		/* PM_INST_CMPL */
+				break;
+			case 0x10000e:	/* PM_PURR */
+				alt[j++] = 0x4000f4;	/* PM_RUN_PURR */
+				break;
+			case 0x4000f4:	/* PM_RUN_PURR */
+				alt[j++] = 0x10000e;	/* PM_PURR */
+				break;
+			}
+		}
+		nalt = j;
+	}
+
+	if (!(flags & PPMU_LIMITED_PMC_OK) && nlim) {
+		/* remove the limited PMC events */
+		j = 0;
+		for (i = 0; i < nalt; ++i) {
+			if (!p6_limited_pmc_event(alt[i])) {
+				alt[j] = alt[i];
+				++j;
+			}
+		}
+		nalt = j;
+	} else if ((flags & PPMU_LIMITED_PMC_REQD) && nlim < nalt) {
+		/* remove all but the limited PMC events */
+		j = 0;
+		for (i = 0; i < nalt; ++i) {
+			if (p6_limited_pmc_event(alt[i])) {
+				alt[j] = alt[i];
+				++j;
+			}
+		}
+		nalt = j;
+	}
+
+	return nalt;
+}
+
+static void p6_disable_pmc(unsigned int pmc, u64 mmcr[])
+{
+	/* Set PMCxSEL to 0 to disable PMCx */
+	if (pmc <= 3)
+		mmcr[1] &= ~(0xffUL << MMCR1_PMCSEL_SH(pmc));
+}
+
+static int power6_generic_events[] = {
+	[PERF_COUNT_HW_CPU_CYCLES]		= 0x1e,
+	[PERF_COUNT_HW_INSTRUCTIONS]		= 2,
+	[PERF_COUNT_HW_CACHE_REFERENCES]	= 0x280030, /* LD_REF_L1 */
+	[PERF_COUNT_HW_CACHE_MISSES]		= 0x30000c, /* LD_MISS_L1 */
+	[PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x410a0,  /* BR_PRED */
+	[PERF_COUNT_HW_BRANCH_MISSES]		= 0x400052, /* BR_MPRED */
+};
+
+#define C(x)	PERF_COUNT_HW_CACHE_##x
+
+/*
+ * Table of generalized cache-related events.
+ * 0 means not supported, -1 means nonsensical, other values
+ * are event codes.
+ * The "DTLB" and "ITLB" events relate to the DERAT and IERAT.
+ */
+static int power6_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
+	[C(L1D)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x80082,	0x80080		},
+		[C(OP_WRITE)] = {	0x80086,	0x80088		},
+		[C(OP_PREFETCH)] = {	0x810a4,	0		},
+	},
+	[C(L1I)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x100056 	},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	0x4008c,	0		},
+	},
+	[C(LL)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x150730,	0x250532	},
+		[C(OP_WRITE)] = {	0x250432,	0x150432	},
+		[C(OP_PREFETCH)] = {	0x810a6,	0		},
+	},
+	[C(DTLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x20000e	},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+	[C(ITLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x420ce		},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+	[C(BPU)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x430e6,	0x400052	},
+		[C(OP_WRITE)] = {	-1,		-1		},
+		[C(OP_PREFETCH)] = {	-1,		-1		},
+	},
+};
+
+struct power_pmu power6_pmu = {
+	.n_counter = 6,
+	.max_alternatives = MAX_ALT,
+	.add_fields = 0x1555,
+	.test_adder = 0x3000,
+	.compute_mmcr = p6_compute_mmcr,
+	.get_constraint = p6_get_constraint,
+	.get_alternatives = p6_get_alternatives,
+	.disable_pmc = p6_disable_pmc,
+	.limited_pmc_event = p6_limited_pmc_event,
+	.flags = PPMU_LIMITED_PMC5_6 | PPMU_ALT_SIPR,
+	.n_generic = ARRAY_SIZE(power6_generic_events),
+	.generic_events = power6_generic_events,
+	.cache_events = &power6_cache_events,
+};
diff --git a/arch/powerpc/kernel/power7-pmu.c b/arch/powerpc/kernel/power7-pmu.c
new file mode 100644
index 00000000000..b3f7d1216ba
--- /dev/null
+++ b/arch/powerpc/kernel/power7-pmu.c
@@ -0,0 +1,357 @@
+/*
+ * Performance counter support for POWER7 processors.
+ *
+ * Copyright 2009 Paul Mackerras, IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+#include <linux/kernel.h>
+#include <linux/perf_counter.h>
+#include <asm/reg.h>
+
+/*
+ * Bits in event code for POWER7
+ */
+#define PM_PMC_SH	16	/* PMC number (1-based) for direct events */
+#define PM_PMC_MSK	0xf
+#define PM_PMC_MSKS	(PM_PMC_MSK << PM_PMC_SH)
+#define PM_UNIT_SH	12	/* TTMMUX number and setting - unit select */
+#define PM_UNIT_MSK	0xf
+#define PM_COMBINE_SH	11	/* Combined event bit */
+#define PM_COMBINE_MSK	1
+#define PM_COMBINE_MSKS	0x800
+#define PM_L2SEL_SH	8	/* L2 event select */
+#define PM_L2SEL_MSK	7
+#define PM_PMCSEL_MSK	0xff
+
+/*
+ * Bits in MMCR1 for POWER7
+ */
+#define MMCR1_TTM0SEL_SH	60
+#define MMCR1_TTM1SEL_SH	56
+#define MMCR1_TTM2SEL_SH	52
+#define MMCR1_TTM3SEL_SH	48
+#define MMCR1_TTMSEL_MSK	0xf
+#define MMCR1_L2SEL_SH		45
+#define MMCR1_L2SEL_MSK		7
+#define MMCR1_PMC1_COMBINE_SH	35
+#define MMCR1_PMC2_COMBINE_SH	34
+#define MMCR1_PMC3_COMBINE_SH	33
+#define MMCR1_PMC4_COMBINE_SH	32
+#define MMCR1_PMC1SEL_SH	24
+#define MMCR1_PMC2SEL_SH	16
+#define MMCR1_PMC3SEL_SH	8
+#define MMCR1_PMC4SEL_SH	0
+#define MMCR1_PMCSEL_SH(n)	(MMCR1_PMC1SEL_SH - (n) * 8)
+#define MMCR1_PMCSEL_MSK	0xff
+
+/*
+ * Bits in MMCRA
+ */
+
+/*
+ * Layout of constraint bits:
+ * 6666555555555544444444443333333333222222222211111111110000000000
+ * 3210987654321098765432109876543210987654321098765432109876543210
+ *                                                 [  ><><><><><><>
+ *                                                  NC P6P5P4P3P2P1
+ *
+ * NC - number of counters
+ *     15: NC error 0x8000
+ *     12-14: number of events needing PMC1-4 0x7000
+ *
+ * P6
+ *     11: P6 error 0x800
+ *     10-11: Count of events needing PMC6
+ *
+ * P1..P5
+ *     0-9: Count of events needing PMC1..PMC5
+ */
+
+static int power7_get_constraint(u64 event, u64 *maskp, u64 *valp)
+{
+	int pmc, sh;
+	u64 mask = 0, value = 0;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	if (pmc) {
+		if (pmc > 6)
+			return -1;
+		sh = (pmc - 1) * 2;
+		mask |= 2 << sh;
+		value |= 1 << sh;
+		if (pmc >= 5 && !(event == 0x500fa || event == 0x600f4))
+			return -1;
+	}
+	if (pmc < 5) {
+		/* need a counter from PMC1-4 set */
+		mask  |= 0x8000;
+		value |= 0x1000;
+	}
+	*maskp = mask;
+	*valp = value;
+	return 0;
+}
+
+#define MAX_ALT	2	/* at most 2 alternatives for any event */
+
+static const unsigned int event_alternatives[][MAX_ALT] = {
+	{ 0x200f2, 0x300f2 },		/* PM_INST_DISP */
+	{ 0x200f4, 0x600f4 },		/* PM_RUN_CYC */
+	{ 0x400fa, 0x500fa },		/* PM_RUN_INST_CMPL */
+};
+
+/*
+ * Scan the alternatives table for a match and return the
+ * index into the alternatives table if found, else -1.
+ */
+static int find_alternative(u64 event)
+{
+	int i, j;
+
+	for (i = 0; i < ARRAY_SIZE(event_alternatives); ++i) {
+		if (event < event_alternatives[i][0])
+			break;
+		for (j = 0; j < MAX_ALT && event_alternatives[i][j]; ++j)
+			if (event == event_alternatives[i][j])
+				return i;
+	}
+	return -1;
+}
+
+static s64 find_alternative_decode(u64 event)
+{
+	int pmc, psel;
+
+	/* this only handles the 4x decode events */
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	psel = event & PM_PMCSEL_MSK;
+	if ((pmc == 2 || pmc == 4) && (psel & ~7) == 0x40)
+		return event - (1 << PM_PMC_SH) + 8;
+	if ((pmc == 1 || pmc == 3) && (psel & ~7) == 0x48)
+		return event + (1 << PM_PMC_SH) - 8;
+	return -1;
+}
+
+static int power7_get_alternatives(u64 event, unsigned int flags, u64 alt[])
+{
+	int i, j, nalt = 1;
+	s64 ae;
+
+	alt[0] = event;
+	nalt = 1;
+	i = find_alternative(event);
+	if (i >= 0) {
+		for (j = 0; j < MAX_ALT; ++j) {
+			ae = event_alternatives[i][j];
+			if (ae && ae != event)
+				alt[nalt++] = ae;
+		}
+	} else {
+		ae = find_alternative_decode(event);
+		if (ae > 0)
+			alt[nalt++] = ae;
+	}
+
+	if (flags & PPMU_ONLY_COUNT_RUN) {
+		/*
+		 * We're only counting in RUN state,
+		 * so PM_CYC is equivalent to PM_RUN_CYC
+		 * and PM_INST_CMPL === PM_RUN_INST_CMPL.
+		 * This doesn't include alternatives that don't provide
+		 * any extra flexibility in assigning PMCs.
+		 */
+		j = nalt;
+		for (i = 0; i < nalt; ++i) {
+			switch (alt[i]) {
+			case 0x1e:	/* PM_CYC */
+				alt[j++] = 0x600f4;	/* PM_RUN_CYC */
+				break;
+			case 0x600f4:	/* PM_RUN_CYC */
+				alt[j++] = 0x1e;
+				break;
+			case 0x2:	/* PM_PPC_CMPL */
+				alt[j++] = 0x500fa;	/* PM_RUN_INST_CMPL */
+				break;
+			case 0x500fa:	/* PM_RUN_INST_CMPL */
+				alt[j++] = 0x2;	/* PM_PPC_CMPL */
+				break;
+			}
+		}
+		nalt = j;
+	}
+
+	return nalt;
+}
+
+/*
+ * Returns 1 if event counts things relating to marked instructions
+ * and thus needs the MMCRA_SAMPLE_ENABLE bit set, or 0 if not.
+ */
+static int power7_marked_instr_event(u64 event)
+{
+	int pmc, psel;
+	int unit;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+	psel = event & PM_PMCSEL_MSK & ~1;	/* trim off edge/level bit */
+	if (pmc >= 5)
+		return 0;
+
+	switch (psel >> 4) {
+	case 2:
+		return pmc == 2 || pmc == 4;
+	case 3:
+		if (psel == 0x3c)
+			return pmc == 1;
+		if (psel == 0x3e)
+			return pmc != 2;
+		return 1;
+	case 4:
+	case 5:
+		return unit == 0xd;
+	case 6:
+		if (psel == 0x64)
+			return pmc >= 3;
+	case 8:
+		return unit == 0xd;
+	}
+	return 0;
+}
+
+static int power7_compute_mmcr(u64 event[], int n_ev,
+			       unsigned int hwc[], u64 mmcr[])
+{
+	u64 mmcr1 = 0;
+	u64 mmcra = 0;
+	unsigned int pmc, unit, combine, l2sel, psel;
+	unsigned int pmc_inuse = 0;
+	int i;
+
+	/* First pass to count resource use */
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		if (pmc) {
+			if (pmc > 6)
+				return -1;
+			if (pmc_inuse & (1 << (pmc - 1)))
+				return -1;
+			pmc_inuse |= 1 << (pmc - 1);
+		}
+	}
+
+	/* Second pass: assign PMCs, set all MMCR1 fields */
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+		combine = (event[i] >> PM_COMBINE_SH) & PM_COMBINE_MSK;
+		l2sel = (event[i] >> PM_L2SEL_SH) & PM_L2SEL_MSK;
+		psel = event[i] & PM_PMCSEL_MSK;
+		if (!pmc) {
+			/* Bus event or any-PMC direct event */
+			for (pmc = 0; pmc < 4; ++pmc) {
+				if (!(pmc_inuse & (1 << pmc)))
+					break;
+			}
+			if (pmc >= 4)
+				return -1;
+			pmc_inuse |= 1 << pmc;
+		} else {
+			/* Direct or decoded event */
+			--pmc;
+		}
+		if (pmc <= 3) {
+			mmcr1 |= (u64) unit << (MMCR1_TTM0SEL_SH - 4 * pmc);
+			mmcr1 |= (u64) combine << (MMCR1_PMC1_COMBINE_SH - pmc);
+			mmcr1 |= psel << MMCR1_PMCSEL_SH(pmc);
+			if (unit == 6)	/* L2 events */
+				mmcr1 |= (u64) l2sel << MMCR1_L2SEL_SH;
+		}
+		if (power7_marked_instr_event(event[i]))
+			mmcra |= MMCRA_SAMPLE_ENABLE;
+		hwc[i] = pmc;
+	}
+
+	/* Return MMCRx values */
+	mmcr[0] = 0;
+	if (pmc_inuse & 1)
+		mmcr[0] = MMCR0_PMC1CE;
+	if (pmc_inuse & 0x3e)
+		mmcr[0] |= MMCR0_PMCjCE;
+	mmcr[1] = mmcr1;
+	mmcr[2] = mmcra;
+	return 0;
+}
+
+static void power7_disable_pmc(unsigned int pmc, u64 mmcr[])
+{
+	if (pmc <= 3)
+		mmcr[1] &= ~(0xffULL << MMCR1_PMCSEL_SH(pmc));
+}
+
+static int power7_generic_events[] = {
+	[PERF_COUNT_CPU_CYCLES] = 0x1e,
+	[PERF_COUNT_INSTRUCTIONS] = 2,
+	[PERF_COUNT_CACHE_REFERENCES] = 0xc880,		/* LD_REF_L1_LSU */
+	[PERF_COUNT_CACHE_MISSES] = 0x400f0,		/* LD_MISS_L1 */
+	[PERF_COUNT_BRANCH_INSTRUCTIONS] = 0x10068,	/* BRU_FIN */
+	[PERF_COUNT_BRANCH_MISSES] = 0x400f6,		/* BR_MPRED */
+};
+
+#define C(x)	PERF_COUNT_HW_CACHE_##x
+
+/*
+ * Table of generalized cache-related events.
+ * 0 means not supported, -1 means nonsensical, other values
+ * are event codes.
+ */
+static int power7_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
+	[C(L1D)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x400f0,	0xc880	},
+		[C(OP_WRITE)] = {	0,		0x300f0	},
+		[C(OP_PREFETCH)] = {	0xd8b8,		0	},
+	},
+	[C(L1I)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x200fc	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	0x408a,		0	},
+	},
+	[C(LL)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x6080,		0x6084	},
+		[C(OP_WRITE)] = {	0x6082,		0x6086	},
+		[C(OP_PREFETCH)] = {	0,		0	},
+	},
+	[C(DTLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x300fc	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+	[C(ITLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x400fc	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+	[C(BPU)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x10068,	0x400f6	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+};
+
+struct power_pmu power7_pmu = {
+	.n_counter = 6,
+	.max_alternatives = MAX_ALT + 1,
+	.add_fields = 0x1555ull,
+	.test_adder = 0x3000ull,
+	.compute_mmcr = power7_compute_mmcr,
+	.get_constraint = power7_get_constraint,
+	.get_alternatives = power7_get_alternatives,
+	.disable_pmc = power7_disable_pmc,
+	.n_generic = ARRAY_SIZE(power7_generic_events),
+	.generic_events = power7_generic_events,
+	.cache_events = &power7_cache_events,
+};
diff --git a/arch/powerpc/kernel/ppc970-pmu.c b/arch/powerpc/kernel/ppc970-pmu.c
new file mode 100644
index 00000000000..ba0a357a89f
--- /dev/null
+++ b/arch/powerpc/kernel/ppc970-pmu.c
@@ -0,0 +1,482 @@
+/*
+ * Performance counter support for PPC970-family processors.
+ *
+ * Copyright 2008-2009 Paul Mackerras, IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+#include <linux/string.h>
+#include <linux/perf_counter.h>
+#include <asm/reg.h>
+
+/*
+ * Bits in event code for PPC970
+ */
+#define PM_PMC_SH	12	/* PMC number (1-based) for direct events */
+#define PM_PMC_MSK	0xf
+#define PM_UNIT_SH	8	/* TTMMUX number and setting - unit select */
+#define PM_UNIT_MSK	0xf
+#define PM_SPCSEL_SH	6
+#define PM_SPCSEL_MSK	3
+#define PM_BYTE_SH	4	/* Byte number of event bus to use */
+#define PM_BYTE_MSK	3
+#define PM_PMCSEL_MSK	0xf
+
+/* Values in PM_UNIT field */
+#define PM_NONE		0
+#define PM_FPU		1
+#define PM_VPU		2
+#define PM_ISU		3
+#define PM_IFU		4
+#define PM_IDU		5
+#define PM_STS		6
+#define PM_LSU0		7
+#define PM_LSU1U	8
+#define PM_LSU1L	9
+#define PM_LASTUNIT	9
+
+/*
+ * Bits in MMCR0 for PPC970
+ */
+#define MMCR0_PMC1SEL_SH	8
+#define MMCR0_PMC2SEL_SH	1
+#define MMCR_PMCSEL_MSK		0x1f
+
+/*
+ * Bits in MMCR1 for PPC970
+ */
+#define MMCR1_TTM0SEL_SH	62
+#define MMCR1_TTM1SEL_SH	59
+#define MMCR1_TTM3SEL_SH	53
+#define MMCR1_TTMSEL_MSK	3
+#define MMCR1_TD_CP_DBG0SEL_SH	50
+#define MMCR1_TD_CP_DBG1SEL_SH	48
+#define MMCR1_TD_CP_DBG2SEL_SH	46
+#define MMCR1_TD_CP_DBG3SEL_SH	44
+#define MMCR1_PMC1_ADDER_SEL_SH	39
+#define MMCR1_PMC2_ADDER_SEL_SH	38
+#define MMCR1_PMC6_ADDER_SEL_SH	37
+#define MMCR1_PMC5_ADDER_SEL_SH	36
+#define MMCR1_PMC8_ADDER_SEL_SH	35
+#define MMCR1_PMC7_ADDER_SEL_SH	34
+#define MMCR1_PMC3_ADDER_SEL_SH	33
+#define MMCR1_PMC4_ADDER_SEL_SH	32
+#define MMCR1_PMC3SEL_SH	27
+#define MMCR1_PMC4SEL_SH	22
+#define MMCR1_PMC5SEL_SH	17
+#define MMCR1_PMC6SEL_SH	12
+#define MMCR1_PMC7SEL_SH	7
+#define MMCR1_PMC8SEL_SH	2
+
+static short mmcr1_adder_bits[8] = {
+	MMCR1_PMC1_ADDER_SEL_SH,
+	MMCR1_PMC2_ADDER_SEL_SH,
+	MMCR1_PMC3_ADDER_SEL_SH,
+	MMCR1_PMC4_ADDER_SEL_SH,
+	MMCR1_PMC5_ADDER_SEL_SH,
+	MMCR1_PMC6_ADDER_SEL_SH,
+	MMCR1_PMC7_ADDER_SEL_SH,
+	MMCR1_PMC8_ADDER_SEL_SH
+};
+
+/*
+ * Bits in MMCRA
+ */
+
+/*
+ * Layout of constraint bits:
+ * 6666555555555544444444443333333333222222222211111111110000000000
+ * 3210987654321098765432109876543210987654321098765432109876543210
+ *               <><><>[  >[  >[  ><  ><  ><  ><  ><><><><><><><><>
+ *               SPT0T1 UC  PS1 PS2 B0  B1  B2  B3 P1P2P3P4P5P6P7P8
+ *
+ * SP - SPCSEL constraint
+ *     48-49: SPCSEL value 0x3_0000_0000_0000
+ *
+ * T0 - TTM0 constraint
+ *     46-47: TTM0SEL value (0=FPU, 2=IFU, 3=VPU) 0xC000_0000_0000
+ *
+ * T1 - TTM1 constraint
+ *     44-45: TTM1SEL value (0=IDU, 3=STS) 0x3000_0000_0000
+ *
+ * UC - unit constraint: can't have all three of FPU|IFU|VPU, ISU, IDU|STS
+ *     43: UC3 error 0x0800_0000_0000
+ *     42: FPU|IFU|VPU events needed 0x0400_0000_0000
+ *     41: ISU events needed 0x0200_0000_0000
+ *     40: IDU|STS events needed 0x0100_0000_0000
+ *
+ * PS1
+ *     39: PS1 error 0x0080_0000_0000
+ *     36-38: count of events needing PMC1/2/5/6 0x0070_0000_0000
+ *
+ * PS2
+ *     35: PS2 error 0x0008_0000_0000
+ *     32-34: count of events needing PMC3/4/7/8 0x0007_0000_0000
+ *
+ * B0
+ *     28-31: Byte 0 event source 0xf000_0000
+ *	      Encoding as for the event code
+ *
+ * B1, B2, B3
+ *     24-27, 20-23, 16-19: Byte 1, 2, 3 event sources
+ *
+ * P1
+ *     15: P1 error 0x8000
+ *     14-15: Count of events needing PMC1
+ *
+ * P2..P8
+ *     0-13: Count of events needing PMC2..PMC8
+ */
+
+static unsigned char direct_marked_event[8] = {
+	(1<<2) | (1<<3),	/* PMC1: PM_MRK_GRP_DISP, PM_MRK_ST_CMPL */
+	(1<<3) | (1<<5),	/* PMC2: PM_THRESH_TIMEO, PM_MRK_BRU_FIN */
+	(1<<3) | (1<<5),	/* PMC3: PM_MRK_ST_CMPL_INT, PM_MRK_VMX_FIN */
+	(1<<4) | (1<<5),	/* PMC4: PM_MRK_GRP_CMPL, PM_MRK_CRU_FIN */
+	(1<<4) | (1<<5),	/* PMC5: PM_GRP_MRK, PM_MRK_GRP_TIMEO */
+	(1<<3) | (1<<4) | (1<<5),
+		/* PMC6: PM_MRK_ST_STS, PM_MRK_FXU_FIN, PM_MRK_GRP_ISSUED */
+	(1<<4) | (1<<5),	/* PMC7: PM_MRK_FPU_FIN, PM_MRK_INST_FIN */
+	(1<<4)			/* PMC8: PM_MRK_LSU_FIN */
+};
+
+/*
+ * Returns 1 if event counts things relating to marked instructions
+ * and thus needs the MMCRA_SAMPLE_ENABLE bit set, or 0 if not.
+ */
+static int p970_marked_instr_event(u64 event)
+{
+	int pmc, psel, unit, byte, bit;
+	unsigned int mask;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	psel = event & PM_PMCSEL_MSK;
+	if (pmc) {
+		if (direct_marked_event[pmc - 1] & (1 << psel))
+			return 1;
+		if (psel == 0)		/* add events */
+			bit = (pmc <= 4)? pmc - 1: 8 - pmc;
+		else if (psel == 7 || psel == 13)	/* decode events */
+			bit = 4;
+		else
+			return 0;
+	} else
+		bit = psel;
+
+	byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+	unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+	mask = 0;
+	switch (unit) {
+	case PM_VPU:
+		mask = 0x4c;		/* byte 0 bits 2,3,6 */
+	case PM_LSU0:
+		/* byte 2 bits 0,2,3,4,6; all of byte 1 */
+		mask = 0x085dff00;
+	case PM_LSU1L:
+		mask = 0x50 << 24;	/* byte 3 bits 4,6 */
+		break;
+	}
+	return (mask >> (byte * 8 + bit)) & 1;
+}
+
+/* Masks and values for using events from the various units */
+static u64 unit_cons[PM_LASTUNIT+1][2] = {
+	[PM_FPU] =   { 0xc80000000000ull, 0x040000000000ull },
+	[PM_VPU] =   { 0xc80000000000ull, 0xc40000000000ull },
+	[PM_ISU] =   { 0x080000000000ull, 0x020000000000ull },
+	[PM_IFU] =   { 0xc80000000000ull, 0x840000000000ull },
+	[PM_IDU] =   { 0x380000000000ull, 0x010000000000ull },
+	[PM_STS] =   { 0x380000000000ull, 0x310000000000ull },
+};
+
+static int p970_get_constraint(u64 event, u64 *maskp, u64 *valp)
+{
+	int pmc, byte, unit, sh, spcsel;
+	u64 mask = 0, value = 0;
+	int grp = -1;
+
+	pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
+	if (pmc) {
+		if (pmc > 8)
+			return -1;
+		sh = (pmc - 1) * 2;
+		mask |= 2 << sh;
+		value |= 1 << sh;
+		grp = ((pmc - 1) >> 1) & 1;
+	}
+	unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
+	if (unit) {
+		if (unit > PM_LASTUNIT)
+			return -1;
+		mask |= unit_cons[unit][0];
+		value |= unit_cons[unit][1];
+		byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
+		/*
+		 * Bus events on bytes 0 and 2 can be counted
+		 * on PMC1/2/5/6; bytes 1 and 3 on PMC3/4/7/8.
+		 */
+		if (!pmc)
+			grp = byte & 1;
+		/* Set byte lane select field */
+		mask  |= 0xfULL << (28 - 4 * byte);
+		value |= (u64)unit << (28 - 4 * byte);
+	}
+	if (grp == 0) {
+		/* increment PMC1/2/5/6 field */
+		mask  |= 0x8000000000ull;
+		value |= 0x1000000000ull;
+	} else if (grp == 1) {
+		/* increment PMC3/4/7/8 field */
+		mask  |= 0x800000000ull;
+		value |= 0x100000000ull;
+	}
+	spcsel = (event >> PM_SPCSEL_SH) & PM_SPCSEL_MSK;
+	if (spcsel) {
+		mask  |= 3ull << 48;
+		value |= (u64)spcsel << 48;
+	}
+	*maskp = mask;
+	*valp = value;
+	return 0;
+}
+
+static int p970_get_alternatives(u64 event, unsigned int flags, u64 alt[])
+{
+	alt[0] = event;
+
+	/* 2 alternatives for LSU empty */
+	if (event == 0x2002 || event == 0x3002) {
+		alt[1] = event ^ 0x1000;
+		return 2;
+	}
+		
+	return 1;
+}
+
+static int p970_compute_mmcr(u64 event[], int n_ev,
+			     unsigned int hwc[], u64 mmcr[])
+{
+	u64 mmcr0 = 0, mmcr1 = 0, mmcra = 0;
+	unsigned int pmc, unit, byte, psel;
+	unsigned int ttm, grp;
+	unsigned int pmc_inuse = 0;
+	unsigned int pmc_grp_use[2];
+	unsigned char busbyte[4];
+	unsigned char unituse[16];
+	unsigned char unitmap[] = { 0, 0<<3, 3<<3, 1<<3, 2<<3, 0|4, 3|4 };
+	unsigned char ttmuse[2];
+	unsigned char pmcsel[8];
+	int i;
+	int spcsel;
+
+	if (n_ev > 8)
+		return -1;
+
+	/* First pass to count resource use */
+	pmc_grp_use[0] = pmc_grp_use[1] = 0;
+	memset(busbyte, 0, sizeof(busbyte));
+	memset(unituse, 0, sizeof(unituse));
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		if (pmc) {
+			if (pmc_inuse & (1 << (pmc - 1)))
+				return -1;
+			pmc_inuse |= 1 << (pmc - 1);
+			/* count 1/2/5/6 vs 3/4/7/8 use */
+			++pmc_grp_use[((pmc - 1) >> 1) & 1];
+		}
+		unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+		byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
+		if (unit) {
+			if (unit > PM_LASTUNIT)
+				return -1;
+			if (!pmc)
+				++pmc_grp_use[byte & 1];
+			if (busbyte[byte] && busbyte[byte] != unit)
+				return -1;
+			busbyte[byte] = unit;
+			unituse[unit] = 1;
+		}
+	}
+	if (pmc_grp_use[0] > 4 || pmc_grp_use[1] > 4)
+		return -1;
+
+	/*
+	 * Assign resources and set multiplexer selects.
+	 *
+	 * PM_ISU can go either on TTM0 or TTM1, but that's the only
+	 * choice we have to deal with.
+	 */
+	if (unituse[PM_ISU] &
+	    (unituse[PM_FPU] | unituse[PM_IFU] | unituse[PM_VPU]))
+		unitmap[PM_ISU] = 2 | 4;	/* move ISU to TTM1 */
+	/* Set TTM[01]SEL fields. */
+	ttmuse[0] = ttmuse[1] = 0;
+	for (i = PM_FPU; i <= PM_STS; ++i) {
+		if (!unituse[i])
+			continue;
+		ttm = unitmap[i];
+		++ttmuse[(ttm >> 2) & 1];
+		mmcr1 |= (u64)(ttm & ~4) << MMCR1_TTM1SEL_SH;
+	}
+	/* Check only one unit per TTMx */
+	if (ttmuse[0] > 1 || ttmuse[1] > 1)
+		return -1;
+
+	/* Set byte lane select fields and TTM3SEL. */
+	for (byte = 0; byte < 4; ++byte) {
+		unit = busbyte[byte];
+		if (!unit)
+			continue;
+		if (unit <= PM_STS)
+			ttm = (unitmap[unit] >> 2) & 1;
+		else if (unit == PM_LSU0)
+			ttm = 2;
+		else {
+			ttm = 3;
+			if (unit == PM_LSU1L && byte >= 2)
+				mmcr1 |= 1ull << (MMCR1_TTM3SEL_SH + 3 - byte);
+		}
+		mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte);
+	}
+
+	/* Second pass: assign PMCs, set PMCxSEL and PMCx_ADDER_SEL fields */
+	memset(pmcsel, 0x8, sizeof(pmcsel));	/* 8 means don't count */
+	for (i = 0; i < n_ev; ++i) {
+		pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
+		unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
+		byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
+		psel = event[i] & PM_PMCSEL_MSK;
+		if (!pmc) {
+			/* Bus event or any-PMC direct event */
+			if (unit)
+				psel |= 0x10 | ((byte & 2) << 2);
+			else
+				psel |= 8;
+			for (pmc = 0; pmc < 8; ++pmc) {
+				if (pmc_inuse & (1 << pmc))
+					continue;
+				grp = (pmc >> 1) & 1;
+				if (unit) {
+					if (grp == (byte & 1))
+						break;
+				} else if (pmc_grp_use[grp] < 4) {
+					++pmc_grp_use[grp];
+					break;
+				}
+			}
+			pmc_inuse |= 1 << pmc;
+		} else {
+			/* Direct event */
+			--pmc;
+			if (psel == 0 && (byte & 2))
+				/* add events on higher-numbered bus */
+				mmcr1 |= 1ull << mmcr1_adder_bits[pmc];
+		}
+		pmcsel[pmc] = psel;
+		hwc[i] = pmc;
+		spcsel = (event[i] >> PM_SPCSEL_SH) & PM_SPCSEL_MSK;
+		mmcr1 |= spcsel;
+		if (p970_marked_instr_event(event[i]))
+			mmcra |= MMCRA_SAMPLE_ENABLE;
+	}
+	for (pmc = 0; pmc < 2; ++pmc)
+		mmcr0 |= pmcsel[pmc] << (MMCR0_PMC1SEL_SH - 7 * pmc);
+	for (; pmc < 8; ++pmc)
+		mmcr1 |= (u64)pmcsel[pmc] << (MMCR1_PMC3SEL_SH - 5 * (pmc - 2));
+	if (pmc_inuse & 1)
+		mmcr0 |= MMCR0_PMC1CE;
+	if (pmc_inuse & 0xfe)
+		mmcr0 |= MMCR0_PMCjCE;
+
+	mmcra |= 0x2000;	/* mark only one IOP per PPC instruction */
+
+	/* Return MMCRx values */
+	mmcr[0] = mmcr0;
+	mmcr[1] = mmcr1;
+	mmcr[2] = mmcra;
+	return 0;
+}
+
+static void p970_disable_pmc(unsigned int pmc, u64 mmcr[])
+{
+	int shift, i;
+
+	if (pmc <= 1) {
+		shift = MMCR0_PMC1SEL_SH - 7 * pmc;
+		i = 0;
+	} else {
+		shift = MMCR1_PMC3SEL_SH - 5 * (pmc - 2);
+		i = 1;
+	}
+	/*
+	 * Setting the PMCxSEL field to 0x08 disables PMC x.
+	 */
+	mmcr[i] = (mmcr[i] & ~(0x1fUL << shift)) | (0x08UL << shift);
+}
+
+static int ppc970_generic_events[] = {
+	[PERF_COUNT_HW_CPU_CYCLES]		= 7,
+	[PERF_COUNT_HW_INSTRUCTIONS]		= 1,
+	[PERF_COUNT_HW_CACHE_REFERENCES]	= 0x8810, /* PM_LD_REF_L1 */
+	[PERF_COUNT_HW_CACHE_MISSES]		= 0x3810, /* PM_LD_MISS_L1 */
+	[PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x431,  /* PM_BR_ISSUED */
+	[PERF_COUNT_HW_BRANCH_MISSES] 		= 0x327,  /* PM_GRP_BR_MPRED */
+};
+
+#define C(x)	PERF_COUNT_HW_CACHE_##x
+
+/*
+ * Table of generalized cache-related events.
+ * 0 means not supported, -1 means nonsensical, other values
+ * are event codes.
+ */
+static int ppc970_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
+	[C(L1D)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x8810,		0x3810	},
+		[C(OP_WRITE)] = {	0x7810,		0x813	},
+		[C(OP_PREFETCH)] = {	0x731,		0	},
+	},
+	[C(L1I)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	0,		0	},
+	},
+	[C(LL)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0	},
+		[C(OP_WRITE)] = {	0,		0	},
+		[C(OP_PREFETCH)] = {	0x733,		0	},
+	},
+	[C(DTLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x704	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+	[C(ITLB)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0,		0x700	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+	[C(BPU)] = {		/* 	RESULT_ACCESS	RESULT_MISS */
+		[C(OP_READ)] = {	0x431,		0x327	},
+		[C(OP_WRITE)] = {	-1,		-1	},
+		[C(OP_PREFETCH)] = {	-1,		-1	},
+	},
+};
+
+struct power_pmu ppc970_pmu = {
+	.n_counter = 8,
+	.max_alternatives = 2,
+	.add_fields = 0x001100005555ull,
+	.test_adder = 0x013300000000ull,
+	.compute_mmcr = p970_compute_mmcr,
+	.get_constraint = p970_get_constraint,
+	.get_alternatives = p970_get_alternatives,
+	.disable_pmc = p970_disable_pmc,
+	.n_generic = ARRAY_SIZE(ppc970_generic_events),
+	.generic_events = ppc970_generic_events,
+	.cache_events = &ppc970_cache_events,
+};
diff --git a/arch/powerpc/mm/fault.c b/arch/powerpc/mm/fault.c
index 76993941cac..5beffc8f481 100644
--- a/arch/powerpc/mm/fault.c
+++ b/arch/powerpc/mm/fault.c
@@ -29,6 +29,7 @@
 #include <linux/module.h>
 #include <linux/kprobes.h>
 #include <linux/kdebug.h>
+#include <linux/perf_counter.h>
 
 #include <asm/firmware.h>
 #include <asm/page.h>
@@ -170,6 +171,8 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 		die("Weird page fault", regs, SIGSEGV);
 	}
 
+	perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
+
 	/* When running in the kernel we expect faults to occur only to
 	 * addresses in user space.  All other faults represent errors in the
 	 * kernel and should generate an OOPS.  Unfortunately, in the case of an
@@ -309,6 +312,8 @@ good_area:
 	}
 	if (ret & VM_FAULT_MAJOR) {
 		current->maj_flt++;
+		perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
+				     regs, address);
 #ifdef CONFIG_PPC_SMLPAR
 		if (firmware_has_feature(FW_FEATURE_CMO)) {
 			preempt_disable();
@@ -316,8 +321,11 @@ good_area:
 			preempt_enable();
 		}
 #endif
-	} else
+	} else {
 		current->min_flt++;
+		perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
+				     regs, address);
+	}
 	up_read(&mm->mmap_sem);
 	return 0;
 
diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype
index 9da795e4933..732ee93a8e9 100644
--- a/arch/powerpc/platforms/Kconfig.cputype
+++ b/arch/powerpc/platforms/Kconfig.cputype
@@ -1,6 +1,7 @@
 config PPC64
 	bool "64-bit kernel"
 	default n
+	select HAVE_PERF_COUNTERS
 	help
 	  This option selects whether a 32-bit or a 64-bit kernel
 	  will be built.
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index aafae3b140d..68f5578fe38 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -739,6 +739,7 @@ config X86_UP_IOAPIC
 config X86_LOCAL_APIC
 	def_bool y
 	depends on X86_64 || SMP || X86_32_NON_STANDARD || X86_UP_APIC
+	select HAVE_PERF_COUNTERS if (!M386 && !M486)
 
 config X86_IO_APIC
 	def_bool y
diff --git a/arch/x86/ia32/ia32entry.S b/arch/x86/ia32/ia32entry.S
index dcef387ddc3..e590261ba05 100644
--- a/arch/x86/ia32/ia32entry.S
+++ b/arch/x86/ia32/ia32entry.S
@@ -825,10 +825,11 @@ ia32_sys_call_table:
 	.quad compat_sys_signalfd4
 	.quad sys_eventfd2
 	.quad sys_epoll_create1
-	.quad sys_dup3			/* 330 */
+	.quad sys_dup3				/* 330 */
 	.quad sys_pipe2
 	.quad sys_inotify_init1
 	.quad compat_sys_preadv
 	.quad compat_sys_pwritev
 	.quad compat_sys_rt_tgsigqueueinfo	/* 335 */
+	.quad sys_perf_counter_open
 ia32_syscall_end:
diff --git a/arch/x86/include/asm/atomic_32.h b/arch/x86/include/asm/atomic_32.h
index 85b46fba422..aff9f1fcdcd 100644
--- a/arch/x86/include/asm/atomic_32.h
+++ b/arch/x86/include/asm/atomic_32.h
@@ -247,5 +247,241 @@ static inline int atomic_add_unless(atomic_t *v, int a, int u)
 #define smp_mb__before_atomic_inc()	barrier()
 #define smp_mb__after_atomic_inc()	barrier()
 
+/* An 64bit atomic type */
+
+typedef struct {
+	unsigned long long counter;
+} atomic64_t;
+
+#define ATOMIC64_INIT(val)	{ (val) }
+
+/**
+ * atomic64_read - read atomic64 variable
+ * @v: pointer of type atomic64_t
+ *
+ * Atomically reads the value of @v.
+ * Doesn't imply a read memory barrier.
+ */
+#define __atomic64_read(ptr)		((ptr)->counter)
+
+static inline unsigned long long
+cmpxchg8b(unsigned long long *ptr, unsigned long long old, unsigned long long new)
+{
+	asm volatile(
+
+		LOCK_PREFIX "cmpxchg8b (%[ptr])\n"
+
+		     :		"=A" (old)
+
+		     : [ptr]	"D" (ptr),
+				"A" (old),
+				"b" (ll_low(new)),
+				"c" (ll_high(new))
+
+		     : "memory");
+
+	return old;
+}
+
+static inline unsigned long long
+atomic64_cmpxchg(atomic64_t *ptr, unsigned long long old_val,
+		 unsigned long long new_val)
+{
+	return cmpxchg8b(&ptr->counter, old_val, new_val);
+}
+
+/**
+ * atomic64_xchg - xchg atomic64 variable
+ * @ptr:      pointer to type atomic64_t
+ * @new_val:  value to assign
+ * @old_val:  old value that was there
+ *
+ * Atomically xchgs the value of @ptr to @new_val and returns
+ * the old value.
+ */
+
+static inline unsigned long long
+atomic64_xchg(atomic64_t *ptr, unsigned long long new_val)
+{
+	unsigned long long old_val;
+
+	do {
+		old_val = atomic_read(ptr);
+	} while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val);
+
+	return old_val;
+}
+
+/**
+ * atomic64_set - set atomic64 variable
+ * @ptr:      pointer to type atomic64_t
+ * @new_val:  value to assign
+ *
+ * Atomically sets the value of @ptr to @new_val.
+ */
+static inline void atomic64_set(atomic64_t *ptr, unsigned long long new_val)
+{
+	atomic64_xchg(ptr, new_val);
+}
+
+/**
+ * atomic64_read - read atomic64 variable
+ * @ptr:      pointer to type atomic64_t
+ *
+ * Atomically reads the value of @ptr and returns it.
+ */
+static inline unsigned long long atomic64_read(atomic64_t *ptr)
+{
+	unsigned long long curr_val;
+
+	do {
+		curr_val = __atomic64_read(ptr);
+	} while (atomic64_cmpxchg(ptr, curr_val, curr_val) != curr_val);
+
+	return curr_val;
+}
+
+/**
+ * atomic64_add_return - add and return
+ * @delta: integer value to add
+ * @ptr:   pointer to type atomic64_t
+ *
+ * Atomically adds @delta to @ptr and returns @delta + *@ptr
+ */
+static inline unsigned long long
+atomic64_add_return(unsigned long long delta, atomic64_t *ptr)
+{
+	unsigned long long old_val, new_val;
+
+	do {
+		old_val = atomic_read(ptr);
+		new_val = old_val + delta;
+
+	} while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val);
+
+	return new_val;
+}
+
+static inline long atomic64_sub_return(unsigned long long delta, atomic64_t *ptr)
+{
+	return atomic64_add_return(-delta, ptr);
+}
+
+static inline long atomic64_inc_return(atomic64_t *ptr)
+{
+	return atomic64_add_return(1, ptr);
+}
+
+static inline long atomic64_dec_return(atomic64_t *ptr)
+{
+	return atomic64_sub_return(1, ptr);
+}
+
+/**
+ * atomic64_add - add integer to atomic64 variable
+ * @delta: integer value to add
+ * @ptr:   pointer to type atomic64_t
+ *
+ * Atomically adds @delta to @ptr.
+ */
+static inline void atomic64_add(unsigned long long delta, atomic64_t *ptr)
+{
+	atomic64_add_return(delta, ptr);
+}
+
+/**
+ * atomic64_sub - subtract the atomic64 variable
+ * @delta: integer value to subtract
+ * @ptr:   pointer to type atomic64_t
+ *
+ * Atomically subtracts @delta from @ptr.
+ */
+static inline void atomic64_sub(unsigned long long delta, atomic64_t *ptr)
+{
+	atomic64_add(-delta, ptr);
+}
+
+/**
+ * atomic64_sub_and_test - subtract value from variable and test result
+ * @delta: integer value to subtract
+ * @ptr:   pointer to type atomic64_t
+ *
+ * Atomically subtracts @delta from @ptr and returns
+ * true if the result is zero, or false for all
+ * other cases.
+ */
+static inline int
+atomic64_sub_and_test(unsigned long long delta, atomic64_t *ptr)
+{
+	unsigned long long old_val = atomic64_sub_return(delta, ptr);
+
+	return old_val == 0;
+}
+
+/**
+ * atomic64_inc - increment atomic64 variable
+ * @ptr: pointer to type atomic64_t
+ *
+ * Atomically increments @ptr by 1.
+ */
+static inline void atomic64_inc(atomic64_t *ptr)
+{
+	atomic64_add(1, ptr);
+}
+
+/**
+ * atomic64_dec - decrement atomic64 variable
+ * @ptr: pointer to type atomic64_t
+ *
+ * Atomically decrements @ptr by 1.
+ */
+static inline void atomic64_dec(atomic64_t *ptr)
+{
+	atomic64_sub(1, ptr);
+}
+
+/**
+ * atomic64_dec_and_test - decrement and test
+ * @ptr: pointer to type atomic64_t
+ *
+ * Atomically decrements @ptr by 1 and
+ * returns true if the result is 0, or false for all other
+ * cases.
+ */
+static inline int atomic64_dec_and_test(atomic64_t *ptr)
+{
+	return atomic64_sub_and_test(1, ptr);
+}
+
+/**
+ * atomic64_inc_and_test - increment and test
+ * @ptr: pointer to type atomic64_t
+ *
+ * Atomically increments @ptr by 1
+ * and returns true if the result is zero, or false for all
+ * other cases.
+ */
+static inline int atomic64_inc_and_test(atomic64_t *ptr)
+{
+	return atomic64_sub_and_test(-1, ptr);
+}
+
+/**
+ * atomic64_add_negative - add and test if negative
+ * @delta: integer value to add
+ * @ptr:   pointer to type atomic64_t
+ *
+ * Atomically adds @delta to @ptr and returns true
+ * if the result is negative, or false when
+ * result is greater than or equal to zero.
+ */
+static inline int
+atomic64_add_negative(unsigned long long delta, atomic64_t *ptr)
+{
+	long long old_val = atomic64_add_return(delta, ptr);
+
+	return old_val < 0;
+}
+
 #include <asm-generic/atomic.h>
 #endif /* _ASM_X86_ATOMIC_32_H */
diff --git a/arch/x86/include/asm/entry_arch.h b/arch/x86/include/asm/entry_arch.h
index c2e6bedaf25..d750a10ccad 100644
--- a/arch/x86/include/asm/entry_arch.h
+++ b/arch/x86/include/asm/entry_arch.h
@@ -49,7 +49,7 @@ BUILD_INTERRUPT(error_interrupt,ERROR_APIC_VECTOR)
 BUILD_INTERRUPT(spurious_interrupt,SPURIOUS_APIC_VECTOR)
 
 #ifdef CONFIG_PERF_COUNTERS
-BUILD_INTERRUPT(perf_counter_interrupt, LOCAL_PERF_VECTOR)
+BUILD_INTERRUPT(perf_pending_interrupt, LOCAL_PENDING_VECTOR)
 #endif
 
 #ifdef CONFIG_X86_MCE_P4THERMAL
diff --git a/arch/x86/include/asm/hardirq.h b/arch/x86/include/asm/hardirq.h
index 37555e52f98..9ebc5c25503 100644
--- a/arch/x86/include/asm/hardirq.h
+++ b/arch/x86/include/asm/hardirq.h
@@ -13,6 +13,8 @@ typedef struct {
 	unsigned int irq_spurious_count;
 #endif
 	unsigned int generic_irqs;	/* arch dependent */
+	unsigned int apic_perf_irqs;
+	unsigned int apic_pending_irqs;
 #ifdef CONFIG_SMP
 	unsigned int irq_resched_count;
 	unsigned int irq_call_count;
diff --git a/arch/x86/include/asm/hw_irq.h b/arch/x86/include/asm/hw_irq.h
index 3bd1777a4c8..6df45f63966 100644
--- a/arch/x86/include/asm/hw_irq.h
+++ b/arch/x86/include/asm/hw_irq.h
@@ -29,6 +29,8 @@
 extern void apic_timer_interrupt(void);
 extern void generic_interrupt(void);
 extern void error_interrupt(void);
+extern void perf_pending_interrupt(void);
+
 extern void spurious_interrupt(void);
 extern void thermal_interrupt(void);
 extern void reschedule_interrupt(void);
diff --git a/arch/x86/include/asm/intel_arch_perfmon.h b/arch/x86/include/asm/intel_arch_perfmon.h
deleted file mode 100644
index fa0fd068bc2..00000000000
--- a/arch/x86/include/asm/intel_arch_perfmon.h
+++ /dev/null
@@ -1,31 +0,0 @@
-#ifndef _ASM_X86_INTEL_ARCH_PERFMON_H
-#define _ASM_X86_INTEL_ARCH_PERFMON_H
-
-#define MSR_ARCH_PERFMON_PERFCTR0		0xc1
-#define MSR_ARCH_PERFMON_PERFCTR1		0xc2
-
-#define MSR_ARCH_PERFMON_EVENTSEL0		0x186
-#define MSR_ARCH_PERFMON_EVENTSEL1		0x187
-
-#define ARCH_PERFMON_EVENTSEL0_ENABLE	(1 << 22)
-#define ARCH_PERFMON_EVENTSEL_INT	(1 << 20)
-#define ARCH_PERFMON_EVENTSEL_OS	(1 << 17)
-#define ARCH_PERFMON_EVENTSEL_USR	(1 << 16)
-
-#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL	(0x3c)
-#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK	(0x00 << 8)
-#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX (0)
-#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT \
-	(1 << (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX))
-
-union cpuid10_eax {
-	struct {
-		unsigned int version_id:8;
-		unsigned int num_counters:8;
-		unsigned int bit_width:8;
-		unsigned int mask_length:8;
-	} split;
-	unsigned int full;
-};
-
-#endif /* _ASM_X86_INTEL_ARCH_PERFMON_H */
diff --git a/arch/x86/include/asm/irq_vectors.h b/arch/x86/include/asm/irq_vectors.h
index 910b5a3d675..e997be98c9b 100644
--- a/arch/x86/include/asm/irq_vectors.h
+++ b/arch/x86/include/asm/irq_vectors.h
@@ -108,14 +108,14 @@
 #define LOCAL_TIMER_VECTOR		0xef
 
 /*
- * Performance monitoring interrupt vector:
+ * Generic system vector for platform specific use
  */
-#define LOCAL_PERF_VECTOR		0xee
+#define GENERIC_INTERRUPT_VECTOR	0xed
 
 /*
- * Generic system vector for platform specific use
+ * Performance monitoring pending work vector:
  */
-#define GENERIC_INTERRUPT_VECTOR	0xed
+#define LOCAL_PENDING_VECTOR		0xec
 
 /*
  * First APIC vector available to drivers: (vectors 0x30-0xee) we
diff --git a/arch/x86/include/asm/perf_counter.h b/arch/x86/include/asm/perf_counter.h
new file mode 100644
index 00000000000..876ed97147b
--- /dev/null
+++ b/arch/x86/include/asm/perf_counter.h
@@ -0,0 +1,100 @@
+#ifndef _ASM_X86_PERF_COUNTER_H
+#define _ASM_X86_PERF_COUNTER_H
+
+/*
+ * Performance counter hw details:
+ */
+
+#define X86_PMC_MAX_GENERIC					8
+#define X86_PMC_MAX_FIXED					3
+
+#define X86_PMC_IDX_GENERIC				        0
+#define X86_PMC_IDX_FIXED				       32
+#define X86_PMC_IDX_MAX					       64
+
+#define MSR_ARCH_PERFMON_PERFCTR0			      0xc1
+#define MSR_ARCH_PERFMON_PERFCTR1			      0xc2
+
+#define MSR_ARCH_PERFMON_EVENTSEL0			     0x186
+#define MSR_ARCH_PERFMON_EVENTSEL1			     0x187
+
+#define ARCH_PERFMON_EVENTSEL0_ENABLE			  (1 << 22)
+#define ARCH_PERFMON_EVENTSEL_INT			  (1 << 20)
+#define ARCH_PERFMON_EVENTSEL_OS			  (1 << 17)
+#define ARCH_PERFMON_EVENTSEL_USR			  (1 << 16)
+
+/*
+ * Includes eventsel and unit mask as well:
+ */
+#define ARCH_PERFMON_EVENT_MASK				    0xffff
+
+#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL		      0x3c
+#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK		(0x00 << 8)
+#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX 		 0
+#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT \
+		(1 << (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX))
+
+#define ARCH_PERFMON_BRANCH_MISSES_RETIRED			 6
+
+/*
+ * Intel "Architectural Performance Monitoring" CPUID
+ * detection/enumeration details:
+ */
+union cpuid10_eax {
+	struct {
+		unsigned int version_id:8;
+		unsigned int num_counters:8;
+		unsigned int bit_width:8;
+		unsigned int mask_length:8;
+	} split;
+	unsigned int full;
+};
+
+union cpuid10_edx {
+	struct {
+		unsigned int num_counters_fixed:4;
+		unsigned int reserved:28;
+	} split;
+	unsigned int full;
+};
+
+
+/*
+ * Fixed-purpose performance counters:
+ */
+
+/*
+ * All 3 fixed-mode PMCs are configured via this single MSR:
+ */
+#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)
+
+/* CPU_CLK_Unhalted.Core: */
+#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)
+
+extern void set_perf_counter_pending(void);
+
+#define clear_perf_counter_pending()	do { } while (0)
+#define test_perf_counter_pending()	(0)
+
+#ifdef CONFIG_PERF_COUNTERS
+extern void init_hw_perf_counters(void);
+extern void perf_counters_lapic_init(void);
+#else
+static inline void init_hw_perf_counters(void)		{ }
+static inline void perf_counters_lapic_init(void)	{ }
+#endif
+
+#endif /* _ASM_X86_PERF_COUNTER_H */
diff --git a/arch/x86/include/asm/unistd_32.h b/arch/x86/include/asm/unistd_32.h
index 708dae61262..732a3070615 100644
--- a/arch/x86/include/asm/unistd_32.h
+++ b/arch/x86/include/asm/unistd_32.h
@@ -341,6 +341,7 @@
 #define __NR_preadv		333
 #define __NR_pwritev		334
 #define __NR_rt_tgsigqueueinfo	335
+#define __NR_perf_counter_open	336
 
 #ifdef __KERNEL__
 
diff --git a/arch/x86/include/asm/unistd_64.h b/arch/x86/include/asm/unistd_64.h
index 4e2b0540440..900e1617e67 100644
--- a/arch/x86/include/asm/unistd_64.h
+++ b/arch/x86/include/asm/unistd_64.h
@@ -659,7 +659,8 @@ __SYSCALL(__NR_preadv, sys_preadv)
 __SYSCALL(__NR_pwritev, sys_pwritev)
 #define __NR_rt_tgsigqueueinfo			297
 __SYSCALL(__NR_rt_tgsigqueueinfo, sys_rt_tgsigqueueinfo)
-
+#define __NR_perf_counter_open			298
+__SYSCALL(__NR_perf_counter_open, sys_perf_counter_open)
 
 #ifndef __NO_STUBS
 #define __ARCH_WANT_OLD_READDIR
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index a4c9cf0bf70..076d3881f3d 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -14,6 +14,7 @@
  *	Mikael Pettersson	:	PM converted to driver model.
  */
 
+#include <linux/perf_counter.h>
 #include <linux/kernel_stat.h>
 #include <linux/mc146818rtc.h>
 #include <linux/acpi_pmtmr.h>
@@ -34,6 +35,7 @@
 #include <linux/smp.h>
 #include <linux/mm.h>
 
+#include <asm/perf_counter.h>
 #include <asm/pgalloc.h>
 #include <asm/atomic.h>
 #include <asm/mpspec.h>
@@ -1187,6 +1189,7 @@ void __cpuinit setup_local_APIC(void)
 		apic_write(APIC_ESR, 0);
 	}
 #endif
+	perf_counters_lapic_init();
 
 	preempt_disable();
 
diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile
index 4e242f9a06e..3efcb2b96a1 100644
--- a/arch/x86/kernel/cpu/Makefile
+++ b/arch/x86/kernel/cpu/Makefile
@@ -1,5 +1,5 @@
 #
-# Makefile for x86-compatible CPU details and quirks
+# Makefile for x86-compatible CPU details, features and quirks
 #
 
 # Don't trace early stages of a secondary CPU boot
@@ -23,11 +23,13 @@ obj-$(CONFIG_CPU_SUP_CENTAUR)		+= centaur.o
 obj-$(CONFIG_CPU_SUP_TRANSMETA_32)	+= transmeta.o
 obj-$(CONFIG_CPU_SUP_UMC_32)		+= umc.o
 
-obj-$(CONFIG_X86_MCE)	+= mcheck/
-obj-$(CONFIG_MTRR)	+= mtrr/
-obj-$(CONFIG_CPU_FREQ)	+= cpufreq/
+obj-$(CONFIG_PERF_COUNTERS)		+= perf_counter.o
 
-obj-$(CONFIG_X86_LOCAL_APIC) += perfctr-watchdog.o
+obj-$(CONFIG_X86_MCE)			+= mcheck/
+obj-$(CONFIG_MTRR)			+= mtrr/
+obj-$(CONFIG_CPU_FREQ)			+= cpufreq/
+
+obj-$(CONFIG_X86_LOCAL_APIC)		+= perfctr-watchdog.o
 
 quiet_cmd_mkcapflags = MKCAP   $@
       cmd_mkcapflags = $(PERL) $(srctree)/$(src)/mkcapflags.pl $< $@
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index b0517aa2bd3..3ffdcfa9abd 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -13,6 +13,7 @@
 #include <linux/io.h>
 
 #include <asm/stackprotector.h>
+#include <asm/perf_counter.h>
 #include <asm/mmu_context.h>
 #include <asm/hypervisor.h>
 #include <asm/processor.h>
@@ -874,6 +875,7 @@ void __init identify_boot_cpu(void)
 #else
 	vgetcpu_set_mode();
 #endif
+	init_hw_perf_counters();
 }
 
 void __cpuinit identify_secondary_cpu(struct cpuinfo_x86 *c)
diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c
new file mode 100644
index 00000000000..895c82e7845
--- /dev/null
+++ b/arch/x86/kernel/cpu/perf_counter.c
@@ -0,0 +1,1704 @@
+/*
+ * Performance counter x86 architecture code
+ *
+ *  Copyright (C) 2008 Thomas Gleixner <tglx@linutronix.de>
+ *  Copyright (C) 2008-2009 Red Hat, Inc., Ingo Molnar
+ *  Copyright (C) 2009 Jaswinder Singh Rajput
+ *  Copyright (C) 2009 Advanced Micro Devices, Inc., Robert Richter
+ *  Copyright (C) 2008-2009 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
+ *
+ *  For licencing details see kernel-base/COPYING
+ */
+
+#include <linux/perf_counter.h>
+#include <linux/capability.h>
+#include <linux/notifier.h>
+#include <linux/hardirq.h>
+#include <linux/kprobes.h>
+#include <linux/module.h>
+#include <linux/kdebug.h>
+#include <linux/sched.h>
+#include <linux/uaccess.h>
+
+#include <asm/apic.h>
+#include <asm/stacktrace.h>
+#include <asm/nmi.h>
+
+static u64 perf_counter_mask __read_mostly;
+
+struct cpu_hw_counters {
+	struct perf_counter	*counters[X86_PMC_IDX_MAX];
+	unsigned long		used_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)];
+	unsigned long		active_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)];
+	unsigned long		interrupts;
+	int			enabled;
+};
+
+/*
+ * struct x86_pmu - generic x86 pmu
+ */
+struct x86_pmu {
+	const char	*name;
+	int		version;
+	int		(*handle_irq)(struct pt_regs *);
+	void		(*disable_all)(void);
+	void		(*enable_all)(void);
+	void		(*enable)(struct hw_perf_counter *, int);
+	void		(*disable)(struct hw_perf_counter *, int);
+	unsigned	eventsel;
+	unsigned	perfctr;
+	u64		(*event_map)(int);
+	u64		(*raw_event)(u64);
+	int		max_events;
+	int		num_counters;
+	int		num_counters_fixed;
+	int		counter_bits;
+	u64		counter_mask;
+	u64		max_period;
+	u64		intel_ctrl;
+};
+
+static struct x86_pmu x86_pmu __read_mostly;
+
+static DEFINE_PER_CPU(struct cpu_hw_counters, cpu_hw_counters) = {
+	.enabled = 1,
+};
+
+/*
+ * Intel PerfMon v3. Used on Core2 and later.
+ */
+static const u64 intel_perfmon_event_map[] =
+{
+  [PERF_COUNT_HW_CPU_CYCLES]		= 0x003c,
+  [PERF_COUNT_HW_INSTRUCTIONS]		= 0x00c0,
+  [PERF_COUNT_HW_CACHE_REFERENCES]	= 0x4f2e,
+  [PERF_COUNT_HW_CACHE_MISSES]		= 0x412e,
+  [PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x00c4,
+  [PERF_COUNT_HW_BRANCH_MISSES]		= 0x00c5,
+  [PERF_COUNT_HW_BUS_CYCLES]		= 0x013c,
+};
+
+static u64 intel_pmu_event_map(int event)
+{
+	return intel_perfmon_event_map[event];
+}
+
+/*
+ * Generalized hw caching related event table, filled
+ * in on a per model basis. A value of 0 means
+ * 'not supported', -1 means 'event makes no sense on
+ * this CPU', any other value means the raw event
+ * ID.
+ */
+
+#define C(x) PERF_COUNT_HW_CACHE_##x
+
+static u64 __read_mostly hw_cache_event_ids
+				[PERF_COUNT_HW_CACHE_MAX]
+				[PERF_COUNT_HW_CACHE_OP_MAX]
+				[PERF_COUNT_HW_CACHE_RESULT_MAX];
+
+static const u64 nehalem_hw_cache_event_ids
+				[PERF_COUNT_HW_CACHE_MAX]
+				[PERF_COUNT_HW_CACHE_OP_MAX]
+				[PERF_COUNT_HW_CACHE_RESULT_MAX] =
+{
+ [ C(L1D) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0f40, /* L1D_CACHE_LD.MESI            */
+		[ C(RESULT_MISS)   ] = 0x0140, /* L1D_CACHE_LD.I_STATE         */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI            */
+		[ C(RESULT_MISS)   ] = 0x0141, /* L1D_CACHE_ST.I_STATE         */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x014e, /* L1D_PREFETCH.REQUESTS        */
+		[ C(RESULT_MISS)   ] = 0x024e, /* L1D_PREFETCH.MISS            */
+	},
+ },
+ [ C(L1I ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0380, /* L1I.READS                    */
+		[ C(RESULT_MISS)   ] = 0x0280, /* L1I.MISSES                   */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0,
+		[ C(RESULT_MISS)   ] = 0x0,
+	},
+ },
+ [ C(LL  ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0324, /* L2_RQSTS.LOADS               */
+		[ C(RESULT_MISS)   ] = 0x0224, /* L2_RQSTS.LD_MISS             */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0c24, /* L2_RQSTS.RFOS                */
+		[ C(RESULT_MISS)   ] = 0x0824, /* L2_RQSTS.RFO_MISS            */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x4f2e, /* LLC Reference                */
+		[ C(RESULT_MISS)   ] = 0x412e, /* LLC Misses                   */
+	},
+ },
+ [ C(DTLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0f40, /* L1D_CACHE_LD.MESI   (alias)  */
+		[ C(RESULT_MISS)   ] = 0x0108, /* DTLB_LOAD_MISSES.ANY         */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI   (alias)  */
+		[ C(RESULT_MISS)   ] = 0x010c, /* MEM_STORE_RETIRED.DTLB_MISS  */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0,
+		[ C(RESULT_MISS)   ] = 0x0,
+	},
+ },
+ [ C(ITLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x01c0, /* INST_RETIRED.ANY_P           */
+		[ C(RESULT_MISS)   ] = 0x20c8, /* ITLB_MISS_RETIRED            */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+ [ C(BPU ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x00c4, /* BR_INST_RETIRED.ALL_BRANCHES */
+		[ C(RESULT_MISS)   ] = 0x03e8, /* BPU_CLEARS.ANY               */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+};
+
+static const u64 core2_hw_cache_event_ids
+				[PERF_COUNT_HW_CACHE_MAX]
+				[PERF_COUNT_HW_CACHE_OP_MAX]
+				[PERF_COUNT_HW_CACHE_RESULT_MAX] =
+{
+ [ C(L1D) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0f40, /* L1D_CACHE_LD.MESI          */
+		[ C(RESULT_MISS)   ] = 0x0140, /* L1D_CACHE_LD.I_STATE       */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI          */
+		[ C(RESULT_MISS)   ] = 0x0141, /* L1D_CACHE_ST.I_STATE       */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x104e, /* L1D_PREFETCH.REQUESTS      */
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(L1I ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0080, /* L1I.READS                  */
+		[ C(RESULT_MISS)   ] = 0x0081, /* L1I.MISSES                 */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(LL  ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x4f29, /* L2_LD.MESI                 */
+		[ C(RESULT_MISS)   ] = 0x4129, /* L2_LD.ISTATE               */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x4f2A, /* L2_ST.MESI                 */
+		[ C(RESULT_MISS)   ] = 0x412A, /* L2_ST.ISTATE               */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(DTLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0f40, /* L1D_CACHE_LD.MESI  (alias) */
+		[ C(RESULT_MISS)   ] = 0x0208, /* DTLB_MISSES.MISS_LD        */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI  (alias) */
+		[ C(RESULT_MISS)   ] = 0x0808, /* DTLB_MISSES.MISS_ST        */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(ITLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x00c0, /* INST_RETIRED.ANY_P         */
+		[ C(RESULT_MISS)   ] = 0x1282, /* ITLBMISSES                 */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+ [ C(BPU ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x00c4, /* BR_INST_RETIRED.ANY        */
+		[ C(RESULT_MISS)   ] = 0x00c5, /* BP_INST_RETIRED.MISPRED    */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+};
+
+static const u64 atom_hw_cache_event_ids
+				[PERF_COUNT_HW_CACHE_MAX]
+				[PERF_COUNT_HW_CACHE_OP_MAX]
+				[PERF_COUNT_HW_CACHE_RESULT_MAX] =
+{
+ [ C(L1D) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x2140, /* L1D_CACHE.LD               */
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x2240, /* L1D_CACHE.ST               */
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(L1I ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0380, /* L1I.READS                  */
+		[ C(RESULT_MISS)   ] = 0x0280, /* L1I.MISSES                 */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(LL  ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x4f29, /* L2_LD.MESI                 */
+		[ C(RESULT_MISS)   ] = 0x4129, /* L2_LD.ISTATE               */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x4f2A, /* L2_ST.MESI                 */
+		[ C(RESULT_MISS)   ] = 0x412A, /* L2_ST.ISTATE               */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(DTLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x2140, /* L1D_CACHE_LD.MESI  (alias) */
+		[ C(RESULT_MISS)   ] = 0x0508, /* DTLB_MISSES.MISS_LD        */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x2240, /* L1D_CACHE_ST.MESI  (alias) */
+		[ C(RESULT_MISS)   ] = 0x0608, /* DTLB_MISSES.MISS_ST        */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(ITLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x00c0, /* INST_RETIRED.ANY_P         */
+		[ C(RESULT_MISS)   ] = 0x0282, /* ITLB.MISSES                */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+ [ C(BPU ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x00c4, /* BR_INST_RETIRED.ANY        */
+		[ C(RESULT_MISS)   ] = 0x00c5, /* BP_INST_RETIRED.MISPRED    */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+};
+
+static u64 intel_pmu_raw_event(u64 event)
+{
+#define CORE_EVNTSEL_EVENT_MASK		0x000000FFULL
+#define CORE_EVNTSEL_UNIT_MASK		0x0000FF00ULL
+#define CORE_EVNTSEL_EDGE_MASK		0x00040000ULL
+#define CORE_EVNTSEL_INV_MASK		0x00800000ULL
+#define CORE_EVNTSEL_COUNTER_MASK	0xFF000000ULL
+
+#define CORE_EVNTSEL_MASK		\
+	(CORE_EVNTSEL_EVENT_MASK |	\
+	 CORE_EVNTSEL_UNIT_MASK  |	\
+	 CORE_EVNTSEL_EDGE_MASK  |	\
+	 CORE_EVNTSEL_INV_MASK  |	\
+	 CORE_EVNTSEL_COUNTER_MASK)
+
+	return event & CORE_EVNTSEL_MASK;
+}
+
+static const u64 amd_0f_hw_cache_event_ids
+				[PERF_COUNT_HW_CACHE_MAX]
+				[PERF_COUNT_HW_CACHE_OP_MAX]
+				[PERF_COUNT_HW_CACHE_RESULT_MAX] =
+{
+ [ C(L1D) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(L1I ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0080, /* Instruction cache fetches  */
+		[ C(RESULT_MISS)   ] = 0x0081, /* Instruction cache misses   */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(LL  ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(DTLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0,
+	},
+ },
+ [ C(ITLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0080, /* Instruction fecthes        */
+		[ C(RESULT_MISS)   ] = 0x0085, /* Instr. fetch ITLB misses   */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+ [ C(BPU ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x00c2, /* Retired Branch Instr.      */
+		[ C(RESULT_MISS)   ] = 0x00c3, /* Retired Mispredicted BI    */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+};
+
+/*
+ * AMD Performance Monitor K7 and later.
+ */
+static const u64 amd_perfmon_event_map[] =
+{
+  [PERF_COUNT_HW_CPU_CYCLES]		= 0x0076,
+  [PERF_COUNT_HW_INSTRUCTIONS]		= 0x00c0,
+  [PERF_COUNT_HW_CACHE_REFERENCES]	= 0x0080,
+  [PERF_COUNT_HW_CACHE_MISSES]		= 0x0081,
+  [PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x00c4,
+  [PERF_COUNT_HW_BRANCH_MISSES]		= 0x00c5,
+};
+
+static u64 amd_pmu_event_map(int event)
+{
+	return amd_perfmon_event_map[event];
+}
+
+static u64 amd_pmu_raw_event(u64 event)
+{
+#define K7_EVNTSEL_EVENT_MASK	0x7000000FFULL
+#define K7_EVNTSEL_UNIT_MASK	0x00000FF00ULL
+#define K7_EVNTSEL_EDGE_MASK	0x000040000ULL
+#define K7_EVNTSEL_INV_MASK	0x000800000ULL
+#define K7_EVNTSEL_COUNTER_MASK	0x0FF000000ULL
+
+#define K7_EVNTSEL_MASK			\
+	(K7_EVNTSEL_EVENT_MASK |	\
+	 K7_EVNTSEL_UNIT_MASK  |	\
+	 K7_EVNTSEL_EDGE_MASK  |	\
+	 K7_EVNTSEL_INV_MASK   |	\
+	 K7_EVNTSEL_COUNTER_MASK)
+
+	return event & K7_EVNTSEL_MASK;
+}
+
+/*
+ * Propagate counter elapsed time into the generic counter.
+ * Can only be executed on the CPU where the counter is active.
+ * Returns the delta events processed.
+ */
+static u64
+x86_perf_counter_update(struct perf_counter *counter,
+			struct hw_perf_counter *hwc, int idx)
+{
+	int shift = 64 - x86_pmu.counter_bits;
+	u64 prev_raw_count, new_raw_count;
+	s64 delta;
+
+	/*
+	 * Careful: an NMI might modify the previous counter value.
+	 *
+	 * Our tactic to handle this is to first atomically read and
+	 * exchange a new raw count - then add that new-prev delta
+	 * count to the generic counter atomically:
+	 */
+again:
+	prev_raw_count = atomic64_read(&hwc->prev_count);
+	rdmsrl(hwc->counter_base + idx, new_raw_count);
+
+	if (atomic64_cmpxchg(&hwc->prev_count, prev_raw_count,
+					new_raw_count) != prev_raw_count)
+		goto again;
+
+	/*
+	 * Now we have the new raw value and have updated the prev
+	 * timestamp already. We can now calculate the elapsed delta
+	 * (counter-)time and add that to the generic counter.
+	 *
+	 * Careful, not all hw sign-extends above the physical width
+	 * of the count.
+	 */
+	delta = (new_raw_count << shift) - (prev_raw_count << shift);
+	delta >>= shift;
+
+	atomic64_add(delta, &counter->count);
+	atomic64_sub(delta, &hwc->period_left);
+
+	return new_raw_count;
+}
+
+static atomic_t active_counters;
+static DEFINE_MUTEX(pmc_reserve_mutex);
+
+static bool reserve_pmc_hardware(void)
+{
+	int i;
+
+	if (nmi_watchdog == NMI_LOCAL_APIC)
+		disable_lapic_nmi_watchdog();
+
+	for (i = 0; i < x86_pmu.num_counters; i++) {
+		if (!reserve_perfctr_nmi(x86_pmu.perfctr + i))
+			goto perfctr_fail;
+	}
+
+	for (i = 0; i < x86_pmu.num_counters; i++) {
+		if (!reserve_evntsel_nmi(x86_pmu.eventsel + i))
+			goto eventsel_fail;
+	}
+
+	return true;
+
+eventsel_fail:
+	for (i--; i >= 0; i--)
+		release_evntsel_nmi(x86_pmu.eventsel + i);
+
+	i = x86_pmu.num_counters;
+
+perfctr_fail:
+	for (i--; i >= 0; i--)
+		release_perfctr_nmi(x86_pmu.perfctr + i);
+
+	if (nmi_watchdog == NMI_LOCAL_APIC)
+		enable_lapic_nmi_watchdog();
+
+	return false;
+}
+
+static void release_pmc_hardware(void)
+{
+	int i;
+
+	for (i = 0; i < x86_pmu.num_counters; i++) {
+		release_perfctr_nmi(x86_pmu.perfctr + i);
+		release_evntsel_nmi(x86_pmu.eventsel + i);
+	}
+
+	if (nmi_watchdog == NMI_LOCAL_APIC)
+		enable_lapic_nmi_watchdog();
+}
+
+static void hw_perf_counter_destroy(struct perf_counter *counter)
+{
+	if (atomic_dec_and_mutex_lock(&active_counters, &pmc_reserve_mutex)) {
+		release_pmc_hardware();
+		mutex_unlock(&pmc_reserve_mutex);
+	}
+}
+
+static inline int x86_pmu_initialized(void)
+{
+	return x86_pmu.handle_irq != NULL;
+}
+
+static inline int
+set_ext_hw_attr(struct hw_perf_counter *hwc, struct perf_counter_attr *attr)
+{
+	unsigned int cache_type, cache_op, cache_result;
+	u64 config, val;
+
+	config = attr->config;
+
+	cache_type = (config >>  0) & 0xff;
+	if (cache_type >= PERF_COUNT_HW_CACHE_MAX)
+		return -EINVAL;
+
+	cache_op = (config >>  8) & 0xff;
+	if (cache_op >= PERF_COUNT_HW_CACHE_OP_MAX)
+		return -EINVAL;
+
+	cache_result = (config >> 16) & 0xff;
+	if (cache_result >= PERF_COUNT_HW_CACHE_RESULT_MAX)
+		return -EINVAL;
+
+	val = hw_cache_event_ids[cache_type][cache_op][cache_result];
+
+	if (val == 0)
+		return -ENOENT;
+
+	if (val == -1)
+		return -EINVAL;
+
+	hwc->config |= val;
+
+	return 0;
+}
+
+/*
+ * Setup the hardware configuration for a given attr_type
+ */
+static int __hw_perf_counter_init(struct perf_counter *counter)
+{
+	struct perf_counter_attr *attr = &counter->attr;
+	struct hw_perf_counter *hwc = &counter->hw;
+	int err;
+
+	if (!x86_pmu_initialized())
+		return -ENODEV;
+
+	err = 0;
+	if (!atomic_inc_not_zero(&active_counters)) {
+		mutex_lock(&pmc_reserve_mutex);
+		if (atomic_read(&active_counters) == 0 && !reserve_pmc_hardware())
+			err = -EBUSY;
+		else
+			atomic_inc(&active_counters);
+		mutex_unlock(&pmc_reserve_mutex);
+	}
+	if (err)
+		return err;
+
+	/*
+	 * Generate PMC IRQs:
+	 * (keep 'enabled' bit clear for now)
+	 */
+	hwc->config = ARCH_PERFMON_EVENTSEL_INT;
+
+	/*
+	 * Count user and OS events unless requested not to.
+	 */
+	if (!attr->exclude_user)
+		hwc->config |= ARCH_PERFMON_EVENTSEL_USR;
+	if (!attr->exclude_kernel)
+		hwc->config |= ARCH_PERFMON_EVENTSEL_OS;
+
+	if (!hwc->sample_period) {
+		hwc->sample_period = x86_pmu.max_period;
+		hwc->last_period = hwc->sample_period;
+		atomic64_set(&hwc->period_left, hwc->sample_period);
+	}
+
+	counter->destroy = hw_perf_counter_destroy;
+
+	/*
+	 * Raw event type provide the config in the event structure
+	 */
+	if (attr->type == PERF_TYPE_RAW) {
+		hwc->config |= x86_pmu.raw_event(attr->config);
+		return 0;
+	}
+
+	if (attr->type == PERF_TYPE_HW_CACHE)
+		return set_ext_hw_attr(hwc, attr);
+
+	if (attr->config >= x86_pmu.max_events)
+		return -EINVAL;
+	/*
+	 * The generic map:
+	 */
+	hwc->config |= x86_pmu.event_map(attr->config);
+
+	return 0;
+}
+
+static void intel_pmu_disable_all(void)
+{
+	wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, 0);
+}
+
+static void amd_pmu_disable_all(void)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	int idx;
+
+	if (!cpuc->enabled)
+		return;
+
+	cpuc->enabled = 0;
+	/*
+	 * ensure we write the disable before we start disabling the
+	 * counters proper, so that amd_pmu_enable_counter() does the
+	 * right thing.
+	 */
+	barrier();
+
+	for (idx = 0; idx < x86_pmu.num_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 &= ~ARCH_PERFMON_EVENTSEL0_ENABLE;
+		wrmsrl(MSR_K7_EVNTSEL0 + idx, val);
+	}
+}
+
+void hw_perf_disable(void)
+{
+	if (!x86_pmu_initialized())
+		return;
+	return x86_pmu.disable_all();
+}
+
+static void intel_pmu_enable_all(void)
+{
+	wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl);
+}
+
+static void amd_pmu_enable_all(void)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	int idx;
+
+	if (cpuc->enabled)
+		return;
+
+	cpuc->enabled = 1;
+	barrier();
+
+	for (idx = 0; idx < x86_pmu.num_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 |= ARCH_PERFMON_EVENTSEL0_ENABLE;
+		wrmsrl(MSR_K7_EVNTSEL0 + idx, val);
+	}
+}
+
+void hw_perf_enable(void)
+{
+	if (!x86_pmu_initialized())
+		return;
+	x86_pmu.enable_all();
+}
+
+static inline u64 intel_pmu_get_status(void)
+{
+	u64 status;
+
+	rdmsrl(MSR_CORE_PERF_GLOBAL_STATUS, status);
+
+	return status;
+}
+
+static inline void intel_pmu_ack_status(u64 ack)
+{
+	wrmsrl(MSR_CORE_PERF_GLOBAL_OVF_CTRL, ack);
+}
+
+static inline void x86_pmu_enable_counter(struct hw_perf_counter *hwc, int idx)
+{
+	int err;
+	err = 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);
+}
+
+static inline void
+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);
+}
+
+static inline void
+intel_pmu_disable_counter(struct hw_perf_counter *hwc, int idx)
+{
+	if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL)) {
+		intel_pmu_disable_fixed(hwc, idx);
+		return;
+	}
+
+	x86_pmu_disable_counter(hwc, idx);
+}
+
+static inline void
+amd_pmu_disable_counter(struct hw_perf_counter *hwc, int idx)
+{
+	x86_pmu_disable_counter(hwc, idx);
+}
+
+static DEFINE_PER_CPU(u64, prev_left[X86_PMC_IDX_MAX]);
+
+/*
+ * Set the next IRQ period, based on the hwc->period_left value.
+ * To be called with the counter disabled in hw:
+ */
+static int
+x86_perf_counter_set_period(struct perf_counter *counter,
+			     struct hw_perf_counter *hwc, int idx)
+{
+	s64 left = atomic64_read(&hwc->period_left);
+	s64 period = hwc->sample_period;
+	int err, ret = 0;
+
+	/*
+	 * If we are way outside a reasoable range then just skip forward:
+	 */
+	if (unlikely(left <= -period)) {
+		left = period;
+		atomic64_set(&hwc->period_left, left);
+		hwc->last_period = period;
+		ret = 1;
+	}
+
+	if (unlikely(left <= 0)) {
+		left += period;
+		atomic64_set(&hwc->period_left, left);
+		hwc->last_period = period;
+		ret = 1;
+	}
+	/*
+	 * Quirk: certain CPUs dont like it if just 1 event is left:
+	 */
+	if (unlikely(left < 2))
+		left = 2;
+
+	if (left > x86_pmu.max_period)
+		left = x86_pmu.max_period;
+
+	per_cpu(prev_left[idx], smp_processor_id()) = left;
+
+	/*
+	 * The hw counter starts counting from this counter offset,
+	 * mark it to be able to extra future deltas:
+	 */
+	atomic64_set(&hwc->prev_count, (u64)-left);
+
+	err = checking_wrmsrl(hwc->counter_base + idx,
+			     (u64)(-left) & x86_pmu.counter_mask);
+
+	return ret;
+}
+
+static inline void
+intel_pmu_enable_fixed(struct hw_perf_counter *hwc, int __idx)
+{
+	int idx = __idx - X86_PMC_IDX_FIXED;
+	u64 ctrl_val, bits, mask;
+	int err;
+
+	/*
+	 * Enable IRQ generation (0x8),
+	 * and enable ring-3 counting (0x2) and ring-0 counting (0x1)
+	 * if requested:
+	 */
+	bits = 0x8ULL;
+	if (hwc->config & ARCH_PERFMON_EVENTSEL_USR)
+		bits |= 0x2;
+	if (hwc->config & ARCH_PERFMON_EVENTSEL_OS)
+		bits |= 0x1;
+	bits <<= (idx * 4);
+	mask = 0xfULL << (idx * 4);
+
+	rdmsrl(hwc->config_base, ctrl_val);
+	ctrl_val &= ~mask;
+	ctrl_val |= bits;
+	err = checking_wrmsrl(hwc->config_base, ctrl_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)) {
+		intel_pmu_enable_fixed(hwc, idx);
+		return;
+	}
+
+	x86_pmu_enable_counter(hwc, idx);
+}
+
+static void amd_pmu_enable_counter(struct hw_perf_counter *hwc, int idx)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+
+	if (cpuc->enabled)
+		x86_pmu_enable_counter(hwc, idx);
+	else
+		x86_pmu_disable_counter(hwc, idx);
+}
+
+static int
+fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc)
+{
+	unsigned int event;
+
+	if (!x86_pmu.num_counters_fixed)
+		return -1;
+
+	event = hwc->config & ARCH_PERFMON_EVENT_MASK;
+
+	if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS)))
+		return X86_PMC_IDX_FIXED_INSTRUCTIONS;
+	if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_CPU_CYCLES)))
+		return X86_PMC_IDX_FIXED_CPU_CYCLES;
+	if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_BUS_CYCLES)))
+		return X86_PMC_IDX_FIXED_BUS_CYCLES;
+
+	return -1;
+}
+
+/*
+ * Find a PMC slot for the freshly enabled / scheduled in counter:
+ */
+static int x86_pmu_enable(struct perf_counter *counter)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	struct hw_perf_counter *hwc = &counter->hw;
+	int idx;
+
+	idx = fixed_mode_idx(counter, hwc);
+	if (idx >= 0) {
+		/*
+		 * Try to get the fixed counter, if that is already taken
+		 * then try to get a generic counter:
+		 */
+		if (test_and_set_bit(idx, cpuc->used_mask))
+			goto try_generic;
+
+		hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL;
+		/*
+		 * We set it so that counter_base + idx in wrmsr/rdmsr maps to
+		 * MSR_ARCH_PERFMON_FIXED_CTR0 ... CTR2:
+		 */
+		hwc->counter_base =
+			MSR_ARCH_PERFMON_FIXED_CTR0 - X86_PMC_IDX_FIXED;
+		hwc->idx = idx;
+	} else {
+		idx = hwc->idx;
+		/* Try to get the previous generic counter again */
+		if (test_and_set_bit(idx, cpuc->used_mask)) {
+try_generic:
+			idx = find_first_zero_bit(cpuc->used_mask,
+						  x86_pmu.num_counters);
+			if (idx == x86_pmu.num_counters)
+				return -EAGAIN;
+
+			set_bit(idx, cpuc->used_mask);
+			hwc->idx = idx;
+		}
+		hwc->config_base  = x86_pmu.eventsel;
+		hwc->counter_base = x86_pmu.perfctr;
+	}
+
+	perf_counters_lapic_init();
+
+	x86_pmu.disable(hwc, idx);
+
+	cpuc->counters[idx] = counter;
+	set_bit(idx, cpuc->active_mask);
+
+	x86_perf_counter_set_period(counter, hwc, idx);
+	x86_pmu.enable(hwc, idx);
+
+	return 0;
+}
+
+static void x86_pmu_unthrottle(struct perf_counter *counter)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	struct hw_perf_counter *hwc = &counter->hw;
+
+	if (WARN_ON_ONCE(hwc->idx >= X86_PMC_IDX_MAX ||
+				cpuc->counters[hwc->idx] != counter))
+		return;
+
+	x86_pmu.enable(hwc, hwc->idx);
+}
+
+void perf_counter_print_debug(void)
+{
+	u64 ctrl, status, overflow, pmc_ctrl, pmc_count, prev_left, fixed;
+	struct cpu_hw_counters *cpuc;
+	unsigned long flags;
+	int cpu, idx;
+
+	if (!x86_pmu.num_counters)
+		return;
+
+	local_irq_save(flags);
+
+	cpu = smp_processor_id();
+	cpuc = &per_cpu(cpu_hw_counters, cpu);
+
+	if (x86_pmu.version >= 2) {
+		rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, ctrl);
+		rdmsrl(MSR_CORE_PERF_GLOBAL_STATUS, status);
+		rdmsrl(MSR_CORE_PERF_GLOBAL_OVF_CTRL, overflow);
+		rdmsrl(MSR_ARCH_PERFMON_FIXED_CTR_CTRL, fixed);
+
+		pr_info("\n");
+		pr_info("CPU#%d: ctrl:       %016llx\n", cpu, ctrl);
+		pr_info("CPU#%d: status:     %016llx\n", cpu, status);
+		pr_info("CPU#%d: overflow:   %016llx\n", cpu, overflow);
+		pr_info("CPU#%d: fixed:      %016llx\n", cpu, fixed);
+	}
+	pr_info("CPU#%d: used:       %016llx\n", cpu, *(u64 *)cpuc->used_mask);
+
+	for (idx = 0; idx < x86_pmu.num_counters; idx++) {
+		rdmsrl(x86_pmu.eventsel + idx, pmc_ctrl);
+		rdmsrl(x86_pmu.perfctr  + idx, pmc_count);
+
+		prev_left = per_cpu(prev_left[idx], cpu);
+
+		pr_info("CPU#%d:   gen-PMC%d ctrl:  %016llx\n",
+			cpu, idx, pmc_ctrl);
+		pr_info("CPU#%d:   gen-PMC%d count: %016llx\n",
+			cpu, idx, pmc_count);
+		pr_info("CPU#%d:   gen-PMC%d left:  %016llx\n",
+			cpu, idx, prev_left);
+	}
+	for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++) {
+		rdmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, pmc_count);
+
+		pr_info("CPU#%d: fixed-PMC%d count: %016llx\n",
+			cpu, idx, pmc_count);
+	}
+	local_irq_restore(flags);
+}
+
+static void x86_pmu_disable(struct perf_counter *counter)
+{
+	struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters);
+	struct hw_perf_counter *hwc = &counter->hw;
+	int idx = hwc->idx;
+
+	/*
+	 * Must be done before we disable, otherwise the nmi handler
+	 * could reenable again:
+	 */
+	clear_bit(idx, cpuc->active_mask);
+	x86_pmu.disable(hwc, idx);
+
+	/*
+	 * Make sure the cleared pointer becomes visible before we
+	 * (potentially) free the counter:
+	 */
+	barrier();
+
+	/*
+	 * Drain the remaining delta count out of a counter
+	 * that we are disabling:
+	 */
+	x86_perf_counter_update(counter, hwc, idx);
+	cpuc->counters[idx] = NULL;
+	clear_bit(idx, cpuc->used_mask);
+}
+
+/*
+ * Save and restart an expired counter. Called by NMI contexts,
+ * so it has to be careful about preempting normal counter ops:
+ */
+static int intel_pmu_save_and_restart(struct perf_counter *counter)
+{
+	struct hw_perf_counter *hwc = &counter->hw;
+	int idx = hwc->idx;
+	int ret;
+
+	x86_perf_counter_update(counter, hwc, idx);
+	ret = x86_perf_counter_set_period(counter, hwc, idx);
+
+	if (counter->state == PERF_COUNTER_STATE_ACTIVE)
+		intel_pmu_enable_counter(hwc, idx);
+
+	return ret;
+}
+
+static void intel_pmu_reset(void)
+{
+	unsigned long flags;
+	int idx;
+
+	if (!x86_pmu.num_counters)
+		return;
+
+	local_irq_save(flags);
+
+	printk("clearing PMU state on CPU#%d\n", smp_processor_id());
+
+	for (idx = 0; idx < x86_pmu.num_counters; idx++) {
+		checking_wrmsrl(x86_pmu.eventsel + idx, 0ull);
+		checking_wrmsrl(x86_pmu.perfctr  + idx, 0ull);
+	}
+	for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++) {
+		checking_wrmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull);
+	}
+
+	local_irq_restore(flags);
+}
+
+
+/*
+ * This handler is triggered by the local APIC, so the APIC IRQ handling
+ * rules apply:
+ */
+static int intel_pmu_handle_irq(struct pt_regs *regs)
+{
+	struct perf_sample_data data;
+	struct cpu_hw_counters *cpuc;
+	int bit, cpu, loops;
+	u64 ack, status;
+
+	data.regs = regs;
+	data.addr = 0;
+
+	cpu = smp_processor_id();
+	cpuc = &per_cpu(cpu_hw_counters, cpu);
+
+	perf_disable();
+	status = intel_pmu_get_status();
+	if (!status) {
+		perf_enable();
+		return 0;
+	}
+
+	loops = 0;
+again:
+	if (++loops > 100) {
+		WARN_ONCE(1, "perfcounters: irq loop stuck!\n");
+		perf_counter_print_debug();
+		intel_pmu_reset();
+		perf_enable();
+		return 1;
+	}
+
+	inc_irq_stat(apic_perf_irqs);
+	ack = status;
+	for_each_bit(bit, (unsigned long *)&status, X86_PMC_IDX_MAX) {
+		struct perf_counter *counter = cpuc->counters[bit];
+
+		clear_bit(bit, (unsigned long *) &status);
+		if (!test_bit(bit, cpuc->active_mask))
+			continue;
+
+		if (!intel_pmu_save_and_restart(counter))
+			continue;
+
+		if (perf_counter_overflow(counter, 1, &data))
+			intel_pmu_disable_counter(&counter->hw, bit);
+	}
+
+	intel_pmu_ack_status(ack);
+
+	/*
+	 * Repeat if there is more work to be done:
+	 */
+	status = intel_pmu_get_status();
+	if (status)
+		goto again;
+
+	perf_enable();
+
+	return 1;
+}
+
+static int amd_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 cpu, idx, handled = 0;
+	u64 val;
+
+	data.regs = regs;
+	data.addr = 0;
+
+	cpu = smp_processor_id();
+	cpuc = &per_cpu(cpu_hw_counters, cpu);
+
+	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))
+			amd_pmu_disable_counter(hwc, idx);
+	}
+
+	if (handled)
+		inc_irq_stat(apic_perf_irqs);
+
+	return handled;
+}
+
+void smp_perf_pending_interrupt(struct pt_regs *regs)
+{
+	irq_enter();
+	ack_APIC_irq();
+	inc_irq_stat(apic_pending_irqs);
+	perf_counter_do_pending();
+	irq_exit();
+}
+
+void set_perf_counter_pending(void)
+{
+	apic->send_IPI_self(LOCAL_PENDING_VECTOR);
+}
+
+void perf_counters_lapic_init(void)
+{
+	if (!x86_pmu_initialized())
+		return;
+
+	/*
+	 * Always use NMI for PMU
+	 */
+	apic_write(APIC_LVTPC, APIC_DM_NMI);
+}
+
+static int __kprobes
+perf_counter_nmi_handler(struct notifier_block *self,
+			 unsigned long cmd, void *__args)
+{
+	struct die_args *args = __args;
+	struct pt_regs *regs;
+
+	if (!atomic_read(&active_counters))
+		return NOTIFY_DONE;
+
+	switch (cmd) {
+	case DIE_NMI:
+	case DIE_NMI_IPI:
+		break;
+
+	default:
+		return NOTIFY_DONE;
+	}
+
+	regs = args->regs;
+
+	apic_write(APIC_LVTPC, APIC_DM_NMI);
+	/*
+	 * Can't rely on the handled return value to say it was our NMI, two
+	 * counters could trigger 'simultaneously' raising two back-to-back NMIs.
+	 *
+	 * If the first NMI handles both, the latter will be empty and daze
+	 * the CPU.
+	 */
+	x86_pmu.handle_irq(regs);
+
+	return NOTIFY_STOP;
+}
+
+static __read_mostly struct notifier_block perf_counter_nmi_notifier = {
+	.notifier_call		= perf_counter_nmi_handler,
+	.next			= NULL,
+	.priority		= 1
+};
+
+static struct x86_pmu intel_pmu = {
+	.name			= "Intel",
+	.handle_irq		= intel_pmu_handle_irq,
+	.disable_all		= intel_pmu_disable_all,
+	.enable_all		= intel_pmu_enable_all,
+	.enable			= intel_pmu_enable_counter,
+	.disable		= intel_pmu_disable_counter,
+	.eventsel		= MSR_ARCH_PERFMON_EVENTSEL0,
+	.perfctr		= MSR_ARCH_PERFMON_PERFCTR0,
+	.event_map		= intel_pmu_event_map,
+	.raw_event		= intel_pmu_raw_event,
+	.max_events		= ARRAY_SIZE(intel_perfmon_event_map),
+	/*
+	 * Intel PMCs cannot be accessed sanely above 32 bit width,
+	 * so we install an artificial 1<<31 period regardless of
+	 * the generic counter period:
+	 */
+	.max_period		= (1ULL << 31) - 1,
+};
+
+static struct x86_pmu amd_pmu = {
+	.name			= "AMD",
+	.handle_irq		= amd_pmu_handle_irq,
+	.disable_all		= amd_pmu_disable_all,
+	.enable_all		= amd_pmu_enable_all,
+	.enable			= amd_pmu_enable_counter,
+	.disable		= amd_pmu_disable_counter,
+	.eventsel		= MSR_K7_EVNTSEL0,
+	.perfctr		= MSR_K7_PERFCTR0,
+	.event_map		= amd_pmu_event_map,
+	.raw_event		= amd_pmu_raw_event,
+	.max_events		= ARRAY_SIZE(amd_perfmon_event_map),
+	.num_counters		= 4,
+	.counter_bits		= 48,
+	.counter_mask		= (1ULL << 48) - 1,
+	/* use highest bit to detect overflow */
+	.max_period		= (1ULL << 47) - 1,
+};
+
+static int intel_pmu_init(void)
+{
+	union cpuid10_edx edx;
+	union cpuid10_eax eax;
+	unsigned int unused;
+	unsigned int ebx;
+	int version;
+
+	if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
+		return -ENODEV;
+
+	/*
+	 * Check whether the Architectural PerfMon supports
+	 * Branch Misses Retired Event or not.
+	 */
+	cpuid(10, &eax.full, &ebx, &unused, &edx.full);
+	if (eax.split.mask_length <= ARCH_PERFMON_BRANCH_MISSES_RETIRED)
+		return -ENODEV;
+
+	version = eax.split.version_id;
+	if (version < 2)
+		return -ENODEV;
+
+	x86_pmu				= intel_pmu;
+	x86_pmu.version			= version;
+	x86_pmu.num_counters		= eax.split.num_counters;
+	x86_pmu.counter_bits		= eax.split.bit_width;
+	x86_pmu.counter_mask		= (1ULL << eax.split.bit_width) - 1;
+
+	/*
+	 * Quirk: v2 perfmon does not report fixed-purpose counters, so
+	 * assume at least 3 counters:
+	 */
+	x86_pmu.num_counters_fixed	= max((int)edx.split.num_counters_fixed, 3);
+
+	rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl);
+
+	/*
+	 * Install the hw-cache-events table:
+	 */
+	switch (boot_cpu_data.x86_model) {
+	case 15: /* original 65 nm celeron/pentium/core2/xeon, "Merom"/"Conroe" */
+	case 22: /* single-core 65 nm celeron/core2solo "Merom-L"/"Conroe-L" */
+	case 23: /* current 45 nm celeron/core2/xeon "Penryn"/"Wolfdale" */
+	case 29: /* six-core 45 nm xeon "Dunnington" */
+		memcpy(hw_cache_event_ids, core2_hw_cache_event_ids,
+		       sizeof(hw_cache_event_ids));
+
+		pr_cont("Core2 events, ");
+		break;
+	default:
+	case 26:
+		memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids,
+		       sizeof(hw_cache_event_ids));
+
+		pr_cont("Nehalem/Corei7 events, ");
+		break;
+	case 28:
+		memcpy(hw_cache_event_ids, atom_hw_cache_event_ids,
+		       sizeof(hw_cache_event_ids));
+
+		pr_cont("Atom events, ");
+		break;
+	}
+	return 0;
+}
+
+static int amd_pmu_init(void)
+{
+	x86_pmu = amd_pmu;
+
+	switch (boot_cpu_data.x86) {
+	case 0x0f:
+	case 0x10:
+	case 0x11:
+		memcpy(hw_cache_event_ids, amd_0f_hw_cache_event_ids,
+		       sizeof(hw_cache_event_ids));
+
+		pr_cont("AMD Family 0f/10/11 events, ");
+		break;
+	}
+	return 0;
+}
+
+void __init init_hw_perf_counters(void)
+{
+	int err;
+
+	pr_info("Performance Counters: ");
+
+	switch (boot_cpu_data.x86_vendor) {
+	case X86_VENDOR_INTEL:
+		err = intel_pmu_init();
+		break;
+	case X86_VENDOR_AMD:
+		err = amd_pmu_init();
+		break;
+	default:
+		return;
+	}
+	if (err != 0) {
+		pr_cont("no PMU driver, software counters only.\n");
+		return;
+	}
+
+	pr_cont("%s PMU driver.\n", x86_pmu.name);
+
+	if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) {
+		x86_pmu.num_counters = X86_PMC_MAX_GENERIC;
+		WARN(1, KERN_ERR "hw perf counters %d > max(%d), clipping!",
+		     x86_pmu.num_counters, X86_PMC_MAX_GENERIC);
+	}
+	perf_counter_mask = (1 << x86_pmu.num_counters) - 1;
+	perf_max_counters = x86_pmu.num_counters;
+
+	if (x86_pmu.num_counters_fixed > X86_PMC_MAX_FIXED) {
+		x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED;
+		WARN(1, KERN_ERR "hw perf counters fixed %d > max(%d), clipping!",
+		     x86_pmu.num_counters_fixed, X86_PMC_MAX_FIXED);
+	}
+
+	perf_counter_mask |=
+		((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED;
+
+	perf_counters_lapic_init();
+	register_die_notifier(&perf_counter_nmi_notifier);
+
+	pr_info("... version:                 %d\n",     x86_pmu.version);
+	pr_info("... bit width:               %d\n",     x86_pmu.counter_bits);
+	pr_info("... generic counters:        %d\n",     x86_pmu.num_counters);
+	pr_info("... value mask:              %016Lx\n", x86_pmu.counter_mask);
+	pr_info("... max period:              %016Lx\n", x86_pmu.max_period);
+	pr_info("... fixed-purpose counters:  %d\n",     x86_pmu.num_counters_fixed);
+	pr_info("... counter mask:            %016Lx\n", perf_counter_mask);
+}
+
+static inline void x86_pmu_read(struct perf_counter *counter)
+{
+	x86_perf_counter_update(counter, &counter->hw, counter->hw.idx);
+}
+
+static const struct pmu pmu = {
+	.enable		= x86_pmu_enable,
+	.disable	= x86_pmu_disable,
+	.read		= x86_pmu_read,
+	.unthrottle	= x86_pmu_unthrottle,
+};
+
+const struct pmu *hw_perf_counter_init(struct perf_counter *counter)
+{
+	int err;
+
+	err = __hw_perf_counter_init(counter);
+	if (err)
+		return ERR_PTR(err);
+
+	return &pmu;
+}
+
+/*
+ * callchain support
+ */
+
+static inline
+void callchain_store(struct perf_callchain_entry *entry, unsigned long ip)
+{
+	if (entry->nr < MAX_STACK_DEPTH)
+		entry->ip[entry->nr++] = ip;
+}
+
+static DEFINE_PER_CPU(struct perf_callchain_entry, irq_entry);
+static DEFINE_PER_CPU(struct perf_callchain_entry, nmi_entry);
+
+
+static void
+backtrace_warning_symbol(void *data, char *msg, unsigned long symbol)
+{
+	/* Ignore warnings */
+}
+
+static void backtrace_warning(void *data, char *msg)
+{
+	/* Ignore warnings */
+}
+
+static int backtrace_stack(void *data, char *name)
+{
+	/* Don't bother with IRQ stacks for now */
+	return -1;
+}
+
+static void backtrace_address(void *data, unsigned long addr, int reliable)
+{
+	struct perf_callchain_entry *entry = data;
+
+	if (reliable)
+		callchain_store(entry, addr);
+}
+
+static const struct stacktrace_ops backtrace_ops = {
+	.warning		= backtrace_warning,
+	.warning_symbol		= backtrace_warning_symbol,
+	.stack			= backtrace_stack,
+	.address		= backtrace_address,
+};
+
+static void
+perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry)
+{
+	unsigned long bp;
+	char *stack;
+	int nr = entry->nr;
+
+	callchain_store(entry, instruction_pointer(regs));
+
+	stack = ((char *)regs + sizeof(struct pt_regs));
+#ifdef CONFIG_FRAME_POINTER
+	bp = frame_pointer(regs);
+#else
+	bp = 0;
+#endif
+
+	dump_trace(NULL, regs, (void *)stack, bp, &backtrace_ops, entry);
+
+	entry->kernel = entry->nr - nr;
+}
+
+
+struct stack_frame {
+	const void __user	*next_fp;
+	unsigned long		return_address;
+};
+
+static int copy_stack_frame(const void __user *fp, struct stack_frame *frame)
+{
+	int ret;
+
+	if (!access_ok(VERIFY_READ, fp, sizeof(*frame)))
+		return 0;
+
+	ret = 1;
+	pagefault_disable();
+	if (__copy_from_user_inatomic(frame, fp, sizeof(*frame)))
+		ret = 0;
+	pagefault_enable();
+
+	return ret;
+}
+
+static void
+perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry)
+{
+	struct stack_frame frame;
+	const void __user *fp;
+	int nr = entry->nr;
+
+	regs = (struct pt_regs *)current->thread.sp0 - 1;
+	fp   = (void __user *)regs->bp;
+
+	callchain_store(entry, regs->ip);
+
+	while (entry->nr < MAX_STACK_DEPTH) {
+		frame.next_fp	     = NULL;
+		frame.return_address = 0;
+
+		if (!copy_stack_frame(fp, &frame))
+			break;
+
+		if ((unsigned long)fp < user_stack_pointer(regs))
+			break;
+
+		callchain_store(entry, frame.return_address);
+		fp = frame.next_fp;
+	}
+
+	entry->user = entry->nr - nr;
+}
+
+static void
+perf_do_callchain(struct pt_regs *regs, struct perf_callchain_entry *entry)
+{
+	int is_user;
+
+	if (!regs)
+		return;
+
+	is_user = user_mode(regs);
+
+	if (!current || current->pid == 0)
+		return;
+
+	if (is_user && current->state != TASK_RUNNING)
+		return;
+
+	if (!is_user)
+		perf_callchain_kernel(regs, entry);
+
+	if (current->mm)
+		perf_callchain_user(regs, entry);
+}
+
+struct perf_callchain_entry *perf_callchain(struct pt_regs *regs)
+{
+	struct perf_callchain_entry *entry;
+
+	if (in_nmi())
+		entry = &__get_cpu_var(nmi_entry);
+	else
+		entry = &__get_cpu_var(irq_entry);
+
+	entry->nr = 0;
+	entry->hv = 0;
+	entry->kernel = 0;
+	entry->user = 0;
+
+	perf_do_callchain(regs, entry);
+
+	return entry;
+}
diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c
index f6c70a164e3..d6f5b9fbde3 100644
--- a/arch/x86/kernel/cpu/perfctr-watchdog.c
+++ b/arch/x86/kernel/cpu/perfctr-watchdog.c
@@ -19,8 +19,8 @@
 #include <linux/nmi.h>
 #include <linux/kprobes.h>
 
-#include <asm/genapic.h>
-#include <asm/intel_arch_perfmon.h>
+#include <asm/apic.h>
+#include <asm/perf_counter.h>
 
 struct nmi_watchdog_ctlblk {
 	unsigned int cccr_msr;
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index 1c17d7c751a..a4742a340d8 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -1012,6 +1012,11 @@ apicinterrupt ERROR_APIC_VECTOR \
 apicinterrupt SPURIOUS_APIC_VECTOR \
 	spurious_interrupt smp_spurious_interrupt
 
+#ifdef CONFIG_PERF_COUNTERS
+apicinterrupt LOCAL_PENDING_VECTOR \
+	perf_pending_interrupt smp_perf_pending_interrupt
+#endif
+
 /*
  * Exception entry points.
  */
diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c
index 9a391bbb8ba..38287b5f116 100644
--- a/arch/x86/kernel/irq.c
+++ b/arch/x86/kernel/irq.c
@@ -62,6 +62,14 @@ static int show_other_interrupts(struct seq_file *p, int prec)
 	for_each_online_cpu(j)
 		seq_printf(p, "%10u ", irq_stats(j)->irq_spurious_count);
 	seq_printf(p, "  Spurious interrupts\n");
+	seq_printf(p, "%*s: ", prec, "CNT");
+	for_each_online_cpu(j)
+		seq_printf(p, "%10u ", irq_stats(j)->apic_perf_irqs);
+	seq_printf(p, "  Performance counter interrupts\n");
+	seq_printf(p, "%*s: ", prec, "PND");
+	for_each_online_cpu(j)
+		seq_printf(p, "%10u ", irq_stats(j)->apic_pending_irqs);
+	seq_printf(p, "  Performance pending work\n");
 #endif
 	if (generic_interrupt_extension) {
 		seq_printf(p, "%*s: ", prec, "PLT");
@@ -165,6 +173,8 @@ u64 arch_irq_stat_cpu(unsigned int cpu)
 #ifdef CONFIG_X86_LOCAL_APIC
 	sum += irq_stats(cpu)->apic_timer_irqs;
 	sum += irq_stats(cpu)->irq_spurious_count;
+	sum += irq_stats(cpu)->apic_perf_irqs;
+	sum += irq_stats(cpu)->apic_pending_irqs;
 #endif
 	if (generic_interrupt_extension)
 		sum += irq_stats(cpu)->generic_irqs;
diff --git a/arch/x86/kernel/irqinit.c b/arch/x86/kernel/irqinit.c
index 2e08b10ad51..267c6624c77 100644
--- a/arch/x86/kernel/irqinit.c
+++ b/arch/x86/kernel/irqinit.c
@@ -181,10 +181,15 @@ static void __init apic_intr_init(void)
 {
 	smp_intr_init();
 
-#ifdef CONFIG_X86_64
+#ifdef CONFIG_X86_THERMAL_VECTOR
 	alloc_intr_gate(THERMAL_APIC_VECTOR, thermal_interrupt);
+#endif
+#ifdef CONFIG_X86_THRESHOLD
 	alloc_intr_gate(THRESHOLD_APIC_VECTOR, threshold_interrupt);
 #endif
+#if defined(CONFIG_X86_NEW_MCE) && defined(CONFIG_X86_LOCAL_APIC)
+	alloc_intr_gate(MCE_SELF_VECTOR, mce_self_interrupt);
+#endif
 
 #if defined(CONFIG_X86_64) || defined(CONFIG_X86_LOCAL_APIC)
 	/* self generated IPI for local APIC timer */
@@ -199,18 +204,10 @@ static void __init apic_intr_init(void)
 
 	/* Performance monitoring interrupts: */
 # ifdef CONFIG_PERF_COUNTERS
-	alloc_intr_gate(LOCAL_PERF_VECTOR, perf_counter_interrupt);
 	alloc_intr_gate(LOCAL_PENDING_VECTOR, perf_pending_interrupt);
 # endif
 
 #endif
-
-#ifdef CONFIG_X86_32
-#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86_MCE_P4THERMAL)
-	/* thermal monitor LVT interrupt */
-	alloc_intr_gate(THERMAL_APIC_VECTOR, thermal_interrupt);
-#endif
-#endif
 }
 
 /**
diff --git a/arch/x86/kernel/signal.c b/arch/x86/kernel/signal.c
index 14425166b8e..0a813b17b17 100644
--- a/arch/x86/kernel/signal.c
+++ b/arch/x86/kernel/signal.c
@@ -6,7 +6,6 @@
  *  2000-06-20  Pentium III FXSR, SSE support by Gareth Hughes
  *  2000-2002   x86-64 support by Andi Kleen
  */
-
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
diff --git a/arch/x86/kernel/syscall_table_32.S b/arch/x86/kernel/syscall_table_32.S
index 734f92c02dd..d51321ddafd 100644
--- a/arch/x86/kernel/syscall_table_32.S
+++ b/arch/x86/kernel/syscall_table_32.S
@@ -335,3 +335,4 @@ ENTRY(sys_call_table)
 	.long sys_preadv
 	.long sys_pwritev
 	.long sys_rt_tgsigqueueinfo	/* 335 */
+	.long sys_perf_counter_open
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index ede024531f8..07d60c870ce 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -942,8 +942,13 @@ void __init trap_init(void)
 #endif
 	set_intr_gate(19, &simd_coprocessor_error);
 
+	/* Reserve all the builtin and the syscall vector: */
+	for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++)
+		set_bit(i, used_vectors);
+
 #ifdef CONFIG_IA32_EMULATION
 	set_system_intr_gate(IA32_SYSCALL_VECTOR, ia32_syscall);
+	set_bit(IA32_SYSCALL_VECTOR, used_vectors);
 #endif
 
 #ifdef CONFIG_X86_32
@@ -960,14 +965,9 @@ void __init trap_init(void)
 	}
 
 	set_system_trap_gate(SYSCALL_VECTOR, &system_call);
+	set_bit(SYSCALL_VECTOR, used_vectors);
 #endif
 
-	/* Reserve all the builtin and the syscall vector: */
-	for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++)
-		set_bit(i, used_vectors);
-
-	set_bit(IA32_SYSCALL_VECTOR, used_vectors);
-
 	/*
 	 * Should be a barrier for any external CPU state:
 	 */
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 5ec7ae36661..c6acc632637 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -10,6 +10,7 @@
 #include <linux/bootmem.h>		/* max_low_pfn			*/
 #include <linux/kprobes.h>		/* __kprobes, ...		*/
 #include <linux/mmiotrace.h>		/* kmmio_handler, ...		*/
+#include <linux/perf_counter.h>		/* perf_swcounter_event		*/
 
 #include <asm/traps.h>			/* dotraplinkage, ...		*/
 #include <asm/pgalloc.h>		/* pgd_*(), ...			*/
@@ -1013,6 +1014,8 @@ do_page_fault(struct pt_regs *regs, unsigned long error_code)
 	if (unlikely(error_code & PF_RSVD))
 		pgtable_bad(regs, error_code, address);
 
+	perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
+
 	/*
 	 * If we're in an interrupt, have no user context or are running
 	 * in an atomic region then we must not take the fault:
@@ -1106,10 +1109,15 @@ good_area:
 		return;
 	}
 
-	if (fault & VM_FAULT_MAJOR)
+	if (fault & VM_FAULT_MAJOR) {
 		tsk->maj_flt++;
-	else
+		perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
+				     regs, address);
+	} else {
 		tsk->min_flt++;
+		perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
+				     regs, address);
+	}
 
 	check_v8086_mode(regs, address, tsk);
 
diff --git a/arch/x86/oprofile/nmi_int.c b/arch/x86/oprofile/nmi_int.c
index 3b285e656e2..b07dd8d0b32 100644
--- a/arch/x86/oprofile/nmi_int.c
+++ b/arch/x86/oprofile/nmi_int.c
@@ -40,8 +40,9 @@ static int profile_exceptions_notify(struct notifier_block *self,
 
 	switch (val) {
 	case DIE_NMI:
-		if (model->check_ctrs(args->regs, &per_cpu(cpu_msrs, cpu)))
-			ret = NOTIFY_STOP;
+	case DIE_NMI_IPI:
+		model->check_ctrs(args->regs, &per_cpu(cpu_msrs, cpu));
+		ret = NOTIFY_STOP;
 		break;
 	default:
 		break;
@@ -134,7 +135,7 @@ static void nmi_cpu_setup(void *dummy)
 static struct notifier_block profile_exceptions_nb = {
 	.notifier_call = profile_exceptions_notify,
 	.next = NULL,
-	.priority = 0
+	.priority = 2
 };
 
 static int nmi_setup(void)
diff --git a/arch/x86/oprofile/op_model_ppro.c b/arch/x86/oprofile/op_model_ppro.c
index 10131fbdaad..4da7230b3d1 100644
--- a/arch/x86/oprofile/op_model_ppro.c
+++ b/arch/x86/oprofile/op_model_ppro.c
@@ -18,7 +18,7 @@
 #include <asm/msr.h>
 #include <asm/apic.h>
 #include <asm/nmi.h>
-#include <asm/intel_arch_perfmon.h>
+#include <asm/perf_counter.h>
 
 #include "op_x86_model.h"
 #include "op_counter.h"
@@ -136,6 +136,13 @@ static int ppro_check_ctrs(struct pt_regs * const regs,
 	u64 val;
 	int i;
 
+	/*
+	 * This can happen if perf counters are in use when
+	 * we steal the die notifier NMI.
+	 */
+	if (unlikely(!reset_value))
+		goto out;
+
 	for (i = 0 ; i < num_counters; ++i) {
 		if (!reset_value[i])
 			continue;
@@ -146,6 +153,7 @@ static int ppro_check_ctrs(struct pt_regs * const regs,
 		}
 	}
 
+out:
 	/* Only P6 based Pentium M need to re-unmask the apic vector but it
 	 * doesn't hurt other P6 variant */
 	apic_write(APIC_LVTPC, apic_read(APIC_LVTPC) & ~APIC_LVT_MASKED);
diff --git a/arch/x86/vdso/vdso32-setup.c b/arch/x86/vdso/vdso32-setup.c
index 1241f118ab5..58bc00f68b1 100644
--- a/arch/x86/vdso/vdso32-setup.c
+++ b/arch/x86/vdso/vdso32-setup.c
@@ -338,6 +338,8 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		}
 	}
 
+	current->mm->context.vdso = (void *)addr;
+
 	if (compat_uses_vma || !compat) {
 		/*
 		 * MAYWRITE to allow gdb to COW and set breakpoints
@@ -358,11 +360,13 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 			goto up_fail;
 	}
 
-	current->mm->context.vdso = (void *)addr;
 	current_thread_info()->sysenter_return =
 		VDSO32_SYMBOL(addr, SYSENTER_RETURN);
 
   up_fail:
+	if (ret)
+		current->mm->context.vdso = NULL;
+
 	up_write(&mm->mmap_sem);
 
 	return ret;
diff --git a/arch/x86/vdso/vma.c b/arch/x86/vdso/vma.c
index cac083386e0..21e1aeb9f3e 100644
--- a/arch/x86/vdso/vma.c
+++ b/arch/x86/vdso/vma.c
@@ -116,15 +116,18 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		goto up_fail;
 	}
 
+	current->mm->context.vdso = (void *)addr;
+
 	ret = install_special_mapping(mm, addr, vdso_size,
 				      VM_READ|VM_EXEC|
 				      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
 				      VM_ALWAYSDUMP,
 				      vdso_pages);
-	if (ret)
+	if (ret) {
+		current->mm->context.vdso = NULL;
 		goto up_fail;
+	}
 
-	current->mm->context.vdso = (void *)addr;
 up_fail:
 	up_write(&mm->mmap_sem);
 	return ret;
author	Linus Torvalds <torvalds@linux-foundation.org>	2009-06-11 14:01:07 -0700
committer	Linus Torvalds <torvalds@linux-foundation.org>	2009-06-11 14:01:07 -0700
commit	8a1ca8cedd108c8e76a6ab34079d0bbb4f244799 (patch)
tree	636c715524f1718599209cc289908ea44b6cb859 /arch
parent	b640f042faa2a2fad6464f259a8afec06e2f6386 (diff)
parent	940010c5a314a7bd9b498593bc6ba1718ac5aec5 (diff)