1 files changed, 310 insertions, 88 deletions

diff --git a/arch/x86/kvm/pmu.c b/arch/x86/kvm/pmu.c
index de1fd7369736..1690d41c1830 100644
--- a/arch/x86/kvm/pmu.c
+++ b/arch/x86/kvm/pmu.c
@@ -9,6 +9,7 @@
  *   Gleb Natapov <gleb@redhat.com>
  *   Wei Huang    <wei@redhat.com>
  */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/types.h>
 #include <linux/kvm_host.h>
@@ -28,9 +29,18 @@
 struct x86_pmu_capability __read_mostly kvm_pmu_cap;
 EXPORT_SYMBOL_GPL(kvm_pmu_cap);
 
-static const struct x86_cpu_id vmx_icl_pebs_cpu[] = {
+/* Precise Distribution of Instructions Retired (PDIR) */
+static const struct x86_cpu_id vmx_pebs_pdir_cpu[] = {
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_D, NULL),
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_X, NULL),
+	/* Instruction-Accurate PDIR (PDIR++) */
+	X86_MATCH_INTEL_FAM6_MODEL(SAPPHIRERAPIDS_X, NULL),
+	{}
+};
+
+/* Precise Distribution (PDist) */
+static const struct x86_cpu_id vmx_pebs_pdist_cpu[] = {
+	X86_MATCH_INTEL_FAM6_MODEL(SAPPHIRERAPIDS_X, NULL),
 	{}
 };
 
@@ -83,7 +93,7 @@ void kvm_pmu_ops_update(const struct kvm_pmu_ops *pmu_ops)
 #undef __KVM_X86_PMU_OP
 }
 
-static inline bool pmc_is_enabled(struct kvm_pmc *pmc)
+static inline bool pmc_is_globally_enabled(struct kvm_pmc *pmc)
 {
 	return static_call(kvm_x86_pmu_pmc_is_enabled)(pmc);
 }
@@ -101,10 +111,6 @@ static inline void __kvm_perf_overflow(struct kvm_pmc *pmc, bool in_pmi)
 	struct kvm_pmu *pmu = pmc_to_pmu(pmc);
 	bool skip_pmi = false;
 
-	/* Ignore counters that have been reprogrammed already. */
-	if (test_and_set_bit(pmc->idx, pmu->reprogram_pmi))
-		return;
-
 	if (pmc->perf_event && pmc->perf_event->attr.precise_ip) {
 		if (!in_pmi) {
 			/*
@@ -122,7 +128,6 @@ static inline void __kvm_perf_overflow(struct kvm_pmc *pmc, bool in_pmi)
 	} else {
 		__set_bit(pmc->idx, (unsigned long *)&pmu->global_status);
 	}
-	kvm_make_request(KVM_REQ_PMU, pmc->vcpu);
 
 	if (!pmc->intr || skip_pmi)
 		return;
@@ -147,12 +152,44 @@ static void kvm_perf_overflow(struct perf_event *perf_event,
 {
 	struct kvm_pmc *pmc = perf_event->overflow_handler_context;
 
+	/*
+	 * Ignore overflow events for counters that are scheduled to be
+	 * reprogrammed, e.g. if a PMI for the previous event races with KVM's
+	 * handling of a related guest WRMSR.
+	 */
+	if (test_and_set_bit(pmc->idx, pmc_to_pmu(pmc)->reprogram_pmi))
+		return;
+
 	__kvm_perf_overflow(pmc, true);
+
+	kvm_make_request(KVM_REQ_PMU, pmc->vcpu);
+}
+
+static u64 pmc_get_pebs_precise_level(struct kvm_pmc *pmc)
+{
+	/*
+	 * For some model specific pebs counters with special capabilities
+	 * (PDIR, PDIR++, PDIST), KVM needs to raise the event precise
+	 * level to the maximum value (currently 3, backwards compatible)
+	 * so that the perf subsystem would assign specific hardware counter
+	 * with that capability for vPMC.
+	 */
+	if ((pmc->idx == 0 && x86_match_cpu(vmx_pebs_pdist_cpu)) ||
+	    (pmc->idx == 32 && x86_match_cpu(vmx_pebs_pdir_cpu)))
+		return 3;
+
+	/*
+	 * The non-zero precision level of guest event makes the ordinary
+	 * guest event becomes a guest PEBS event and triggers the host
+	 * PEBS PMI handler to determine whether the PEBS overflow PMI
+	 * comes from the host counters or the guest.
+	 */
+	return 1;
 }
 
-static void pmc_reprogram_counter(struct kvm_pmc *pmc, u32 type,
-				  u64 config, bool exclude_user,
-				  bool exclude_kernel, bool intr)
+static int pmc_reprogram_counter(struct kvm_pmc *pmc, u32 type, u64 config,
+				 bool exclude_user, bool exclude_kernel,
+				 bool intr)
 {
 	struct kvm_pmu *pmu = pmc_to_pmu(pmc);
 	struct perf_event *event;
@@ -181,22 +218,12 @@ static void pmc_reprogram_counter(struct kvm_pmc *pmc, u32 type,
 	}
 	if (pebs) {
 		/*
-		 * The non-zero precision level of guest event makes the ordinary
-		 * guest event becomes a guest PEBS event and triggers the host
-		 * PEBS PMI handler to determine whether the PEBS overflow PMI
-		 * comes from the host counters or the guest.
-		 *
 		 * For most PEBS hardware events, the difference in the software
 		 * precision levels of guest and host PEBS events will not affect
 		 * the accuracy of the PEBS profiling result, because the "event IP"
 		 * in the PEBS record is calibrated on the guest side.
-		 *
-		 * On Icelake everything is fine. Other hardware (GLC+, TNT+) that
-		 * could possibly care here is unsupported and needs changes.
 		 */
-		attr.precise_ip = 1;
-		if (x86_match_cpu(vmx_icl_pebs_cpu) && pmc->idx == 32)
-			attr.precise_ip = 3;
+		attr.precise_ip = pmc_get_pebs_precise_level(pmc);
 	}
 
 	event = perf_event_create_kernel_counter(&attr, -1, current,
@@ -204,14 +231,14 @@ static void pmc_reprogram_counter(struct kvm_pmc *pmc, u32 type,
 	if (IS_ERR(event)) {
 		pr_debug_ratelimited("kvm_pmu: event creation failed %ld for pmc->idx = %d\n",
 			    PTR_ERR(event), pmc->idx);
-		return;
+		return PTR_ERR(event);
 	}
 
 	pmc->perf_event = event;
 	pmc_to_pmu(pmc)->event_count++;
-	clear_bit(pmc->idx, pmc_to_pmu(pmc)->reprogram_pmi);
 	pmc->is_paused = false;
 	pmc->intr = intr || pebs;
+	return 0;
 }
 
 static void pmc_pause_counter(struct kvm_pmc *pmc)
@@ -233,7 +260,8 @@ static bool pmc_resume_counter(struct kvm_pmc *pmc)
 		return false;
 
 	/* recalibrate sample period and check if it's accepted by perf core */
-	if (perf_event_period(pmc->perf_event,
+	if (is_sampling_event(pmc->perf_event) &&
+	    perf_event_period(pmc->perf_event,
 			      get_sample_period(pmc, pmc->counter)))
 		return false;
 
@@ -245,55 +273,140 @@ static bool pmc_resume_counter(struct kvm_pmc *pmc)
 	perf_event_enable(pmc->perf_event);
 	pmc->is_paused = false;
 
-	clear_bit(pmc->idx, (unsigned long *)&pmc_to_pmu(pmc)->reprogram_pmi);
 	return true;
 }
 
-static int cmp_u64(const void *pa, const void *pb)
+static int filter_cmp(const void *pa, const void *pb, u64 mask)
 {
-	u64 a = *(u64 *)pa;
-	u64 b = *(u64 *)pb;
+	u64 a = *(u64 *)pa & mask;
+	u64 b = *(u64 *)pb & mask;
 
 	return (a > b) - (a < b);
 }
 
+
+static int filter_sort_cmp(const void *pa, const void *pb)
+{
+	return filter_cmp(pa, pb, (KVM_PMU_MASKED_ENTRY_EVENT_SELECT |
+				   KVM_PMU_MASKED_ENTRY_EXCLUDE));
+}
+
+/*
+ * For the event filter, searching is done on the 'includes' list and
+ * 'excludes' list separately rather than on the 'events' list (which
+ * has both).  As a result the exclude bit can be ignored.
+ */
+static int filter_event_cmp(const void *pa, const void *pb)
+{
+	return filter_cmp(pa, pb, (KVM_PMU_MASKED_ENTRY_EVENT_SELECT));
+}
+
+static int find_filter_index(u64 *events, u64 nevents, u64 key)
+{
+	u64 *fe = bsearch(&key, events, nevents, sizeof(events[0]),
+			  filter_event_cmp);
+
+	if (!fe)
+		return -1;
+
+	return fe - events;
+}
+
+static bool is_filter_entry_match(u64 filter_event, u64 umask)
+{
+	u64 mask = filter_event >> (KVM_PMU_MASKED_ENTRY_UMASK_MASK_SHIFT - 8);
+	u64 match = filter_event & KVM_PMU_MASKED_ENTRY_UMASK_MATCH;
+
+	BUILD_BUG_ON((KVM_PMU_ENCODE_MASKED_ENTRY(0, 0xff, 0, false) >>
+		     (KVM_PMU_MASKED_ENTRY_UMASK_MASK_SHIFT - 8)) !=
+		     ARCH_PERFMON_EVENTSEL_UMASK);
+
+	return (umask & mask) == match;
+}
+
+static bool filter_contains_match(u64 *events, u64 nevents, u64 eventsel)
+{
+	u64 event_select = eventsel & kvm_pmu_ops.EVENTSEL_EVENT;
+	u64 umask = eventsel & ARCH_PERFMON_EVENTSEL_UMASK;
+	int i, index;
+
+	index = find_filter_index(events, nevents, event_select);
+	if (index < 0)
+		return false;
+
+	/*
+	 * Entries are sorted by the event select.  Walk the list in both
+	 * directions to process all entries with the targeted event select.
+	 */
+	for (i = index; i < nevents; i++) {
+		if (filter_event_cmp(&events[i], &event_select))
+			break;
+
+		if (is_filter_entry_match(events[i], umask))
+			return true;
+	}
+
+	for (i = index - 1; i >= 0; i--) {
+		if (filter_event_cmp(&events[i], &event_select))
+			break;
+
+		if (is_filter_entry_match(events[i], umask))
+			return true;
+	}
+
+	return false;
+}
+
+static bool is_gp_event_allowed(struct kvm_x86_pmu_event_filter *f,
+				u64 eventsel)
+{
+	if (filter_contains_match(f->includes, f->nr_includes, eventsel) &&
+	    !filter_contains_match(f->excludes, f->nr_excludes, eventsel))
+		return f->action == KVM_PMU_EVENT_ALLOW;
+
+	return f->action == KVM_PMU_EVENT_DENY;
+}
+
+static bool is_fixed_event_allowed(struct kvm_x86_pmu_event_filter *filter,
+				   int idx)
+{
+	int fixed_idx = idx - INTEL_PMC_IDX_FIXED;
+
+	if (filter->action == KVM_PMU_EVENT_DENY &&
+	    test_bit(fixed_idx, (ulong *)&filter->fixed_counter_bitmap))
+		return false;
+	if (filter->action == KVM_PMU_EVENT_ALLOW &&
+	    !test_bit(fixed_idx, (ulong *)&filter->fixed_counter_bitmap))
+		return false;
+
+	return true;
+}
+
 static bool check_pmu_event_filter(struct kvm_pmc *pmc)
 {
-	struct kvm_pmu_event_filter *filter;
+	struct kvm_x86_pmu_event_filter *filter;
 	struct kvm *kvm = pmc->vcpu->kvm;
-	bool allow_event = true;
-	__u64 key;
-	int idx;
 
 	if (!static_call(kvm_x86_pmu_hw_event_available)(pmc))
 		return false;
 
 	filter = srcu_dereference(kvm->arch.pmu_event_filter, &kvm->srcu);
 	if (!filter)
-		goto out;
+		return true;
 
-	if (pmc_is_gp(pmc)) {
-		key = pmc->eventsel & AMD64_RAW_EVENT_MASK_NB;
-		if (bsearch(&key, filter->events, filter->nevents,
-			    sizeof(__u64), cmp_u64))
-			allow_event = filter->action == KVM_PMU_EVENT_ALLOW;
-		else
-			allow_event = filter->action == KVM_PMU_EVENT_DENY;
-	} else {
-		idx = pmc->idx - INTEL_PMC_IDX_FIXED;
-		if (filter->action == KVM_PMU_EVENT_DENY &&
-		    test_bit(idx, (ulong *)&filter->fixed_counter_bitmap))
-			allow_event = false;
-		if (filter->action == KVM_PMU_EVENT_ALLOW &&
-		    !test_bit(idx, (ulong *)&filter->fixed_counter_bitmap))
-			allow_event = false;
-	}
+	if (pmc_is_gp(pmc))
+		return is_gp_event_allowed(filter, pmc->eventsel);
 
-out:
-	return allow_event;
+	return is_fixed_event_allowed(filter, pmc->idx);
 }
 
-void reprogram_counter(struct kvm_pmc *pmc)
+static bool pmc_event_is_allowed(struct kvm_pmc *pmc)
+{
+	return pmc_is_globally_enabled(pmc) && pmc_speculative_in_use(pmc) &&
+	       check_pmu_event_filter(pmc);
+}
+
+static void reprogram_counter(struct kvm_pmc *pmc)
 {
 	struct kvm_pmu *pmu = pmc_to_pmu(pmc);
 	u64 eventsel = pmc->eventsel;
@@ -302,11 +415,11 @@ void reprogram_counter(struct kvm_pmc *pmc)
 
 	pmc_pause_counter(pmc);
 
-	if (!pmc_speculative_in_use(pmc) || !pmc_is_enabled(pmc))
-		return;
+	if (!pmc_event_is_allowed(pmc))
+		goto reprogram_complete;
 
-	if (!check_pmu_event_filter(pmc))
-		return;
+	if (pmc->counter < pmc->prev_counter)
+		__kvm_perf_overflow(pmc, false);
 
 	if (eventsel & ARCH_PERFMON_EVENTSEL_PIN_CONTROL)
 		printk_once("kvm pmu: pin control bit is ignored\n");
@@ -324,18 +437,29 @@ void reprogram_counter(struct kvm_pmc *pmc)
 	}
 
 	if (pmc->current_config == new_config && pmc_resume_counter(pmc))
-		return;
+		goto reprogram_complete;
 
 	pmc_release_perf_event(pmc);
 
 	pmc->current_config = new_config;
-	pmc_reprogram_counter(pmc, PERF_TYPE_RAW,
-			      (eventsel & pmu->raw_event_mask),
-			      !(eventsel & ARCH_PERFMON_EVENTSEL_USR),
-			      !(eventsel & ARCH_PERFMON_EVENTSEL_OS),
-			      eventsel & ARCH_PERFMON_EVENTSEL_INT);
+
+	/*
+	 * If reprogramming fails, e.g. due to contention, leave the counter's
+	 * regprogram bit set, i.e. opportunistically try again on the next PMU
+	 * refresh.  Don't make a new request as doing so can stall the guest
+	 * if reprogramming repeatedly fails.
+	 */
+	if (pmc_reprogram_counter(pmc, PERF_TYPE_RAW,
+				  (eventsel & pmu->raw_event_mask),
+				  !(eventsel & ARCH_PERFMON_EVENTSEL_USR),
+				  !(eventsel & ARCH_PERFMON_EVENTSEL_OS),
+				  eventsel & ARCH_PERFMON_EVENTSEL_INT))
+		return;
+
+reprogram_complete:
+	clear_bit(pmc->idx, (unsigned long *)&pmc_to_pmu(pmc)->reprogram_pmi);
+	pmc->prev_counter = 0;
 }
-EXPORT_SYMBOL_GPL(reprogram_counter);
 
 void kvm_pmu_handle_event(struct kvm_vcpu *vcpu)
 {
@@ -345,10 +469,11 @@ void kvm_pmu_handle_event(struct kvm_vcpu *vcpu)
 	for_each_set_bit(bit, pmu->reprogram_pmi, X86_PMC_IDX_MAX) {
 		struct kvm_pmc *pmc = static_call(kvm_x86_pmu_pmc_idx_to_pmc)(pmu, bit);
 
-		if (unlikely(!pmc || !pmc->perf_event)) {
+		if (unlikely(!pmc)) {
 			clear_bit(bit, pmu->reprogram_pmi);
 			continue;
 		}
+
 		reprogram_counter(pmc);
 	}
 
@@ -418,9 +543,9 @@ int kvm_pmu_rdpmc(struct kvm_vcpu *vcpu, unsigned idx, u64 *data)
 	if (!pmc)
 		return 1;
 
-	if (!(kvm_read_cr4(vcpu) & X86_CR4_PCE) &&
+	if (!kvm_is_cr4_bit_set(vcpu, X86_CR4_PCE) &&
 	    (static_call(kvm_x86_get_cpl)(vcpu) != 0) &&
-	    (kvm_read_cr0(vcpu) & X86_CR0_PE))
+	    kvm_is_cr0_bit_set(vcpu, X86_CR0_PE))
 		return 1;
 
 	*data = pmc_read_counter(pmc) & mask;
@@ -467,6 +592,10 @@ int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
  */
 void kvm_pmu_refresh(struct kvm_vcpu *vcpu)
 {
+	if (KVM_BUG_ON(kvm_vcpu_has_run(vcpu), vcpu->kvm))
+		return;
+
+	bitmap_zero(vcpu_to_pmu(vcpu)->all_valid_pmc_idx, X86_PMC_IDX_MAX);
 	static_call(kvm_x86_pmu_refresh)(vcpu);
 }
 
@@ -522,14 +651,9 @@ void kvm_pmu_destroy(struct kvm_vcpu *vcpu)
 
 static void kvm_pmu_incr_counter(struct kvm_pmc *pmc)
 {
-	u64 prev_count;
-
-	prev_count = pmc->counter;
+	pmc->prev_counter = pmc->counter;
 	pmc->counter = (pmc->counter + 1) & pmc_bitmask(pmc);
-
-	reprogram_counter(pmc);
-	if (pmc->counter < prev_count)
-		__kvm_perf_overflow(pmc, false);
+	kvm_pmu_request_counter_reprogram(pmc);
 }
 
 static inline bool eventsel_match_perf_hw_id(struct kvm_pmc *pmc,
@@ -542,12 +666,15 @@ static inline bool eventsel_match_perf_hw_id(struct kvm_pmc *pmc,
 static inline bool cpl_is_matched(struct kvm_pmc *pmc)
 {
 	bool select_os, select_user;
-	u64 config = pmc->current_config;
+	u64 config;
 
 	if (pmc_is_gp(pmc)) {
+		config = pmc->eventsel;
 		select_os = config & ARCH_PERFMON_EVENTSEL_OS;
 		select_user = config & ARCH_PERFMON_EVENTSEL_USR;
 	} else {
+		config = fixed_ctrl_field(pmc_to_pmu(pmc)->fixed_ctr_ctrl,
+					  pmc->idx - INTEL_PMC_IDX_FIXED);
 		select_os = config & 0x1;
 		select_user = config & 0x2;
 	}
@@ -564,7 +691,7 @@ void kvm_pmu_trigger_event(struct kvm_vcpu *vcpu, u64 perf_hw_id)
 	for_each_set_bit(i, pmu->all_valid_pmc_idx, X86_PMC_IDX_MAX) {
 		pmc = static_call(kvm_x86_pmu_pmc_idx_to_pmc)(pmu, i);
 
-		if (!pmc || !pmc_is_enabled(pmc) || !pmc_speculative_in_use(pmc))
+		if (!pmc || !pmc_event_is_allowed(pmc))
 			continue;
 
 		/* Ignore checks for edge detect, pin control, invert and CMASK bits */
@@ -574,48 +701,143 @@ void kvm_pmu_trigger_event(struct kvm_vcpu *vcpu, u64 perf_hw_id)
 }
 EXPORT_SYMBOL_GPL(kvm_pmu_trigger_event);
 
+static bool is_masked_filter_valid(const struct kvm_x86_pmu_event_filter *filter)
+{
+	u64 mask = kvm_pmu_ops.EVENTSEL_EVENT |
+		   KVM_PMU_MASKED_ENTRY_UMASK_MASK |
+		   KVM_PMU_MASKED_ENTRY_UMASK_MATCH |
+		   KVM_PMU_MASKED_ENTRY_EXCLUDE;
+	int i;
+
+	for (i = 0; i < filter->nevents; i++) {
+		if (filter->events[i] & ~mask)
+			return false;
+	}
+
+	return true;
+}
+
+static void convert_to_masked_filter(struct kvm_x86_pmu_event_filter *filter)
+{
+	int i, j;
+
+	for (i = 0, j = 0; i < filter->nevents; i++) {
+		/*
+		 * Skip events that are impossible to match against a guest
+		 * event.  When filtering, only the event select + unit mask
+		 * of the guest event is used.  To maintain backwards
+		 * compatibility, impossible filters can't be rejected :-(
+		 */
+		if (filter->events[i] & ~(kvm_pmu_ops.EVENTSEL_EVENT |
+					  ARCH_PERFMON_EVENTSEL_UMASK))
+			continue;
+		/*
+		 * Convert userspace events to a common in-kernel event so
+		 * only one code path is needed to support both events.  For
+		 * the in-kernel events use masked events because they are
+		 * flexible enough to handle both cases.  To convert to masked
+		 * events all that's needed is to add an "all ones" umask_mask,
+		 * (unmasked filter events don't support EXCLUDE).
+		 */
+		filter->events[j++] = filter->events[i] |
+				      (0xFFULL << KVM_PMU_MASKED_ENTRY_UMASK_MASK_SHIFT);
+	}
+
+	filter->nevents = j;
+}
+
+static int prepare_filter_lists(struct kvm_x86_pmu_event_filter *filter)
+{
+	int i;
+
+	if (!(filter->flags & KVM_PMU_EVENT_FLAG_MASKED_EVENTS))
+		convert_to_masked_filter(filter);
+	else if (!is_masked_filter_valid(filter))
+		return -EINVAL;
+
+	/*
+	 * Sort entries by event select and includes vs. excludes so that all
+	 * entries for a given event select can be processed efficiently during
+	 * filtering.  The EXCLUDE flag uses a more significant bit than the
+	 * event select, and so the sorted list is also effectively split into
+	 * includes and excludes sub-lists.
+	 */
+	sort(&filter->events, filter->nevents, sizeof(filter->events[0]),
+	     filter_sort_cmp, NULL);
+
+	i = filter->nevents;
+	/* Find the first EXCLUDE event (only supported for masked events). */
+	if (filter->flags & KVM_PMU_EVENT_FLAG_MASKED_EVENTS) {
+		for (i = 0; i < filter->nevents; i++) {
+			if (filter->events[i] & KVM_PMU_MASKED_ENTRY_EXCLUDE)
+				break;
+		}
+	}
+
+	filter->nr_includes = i;
+	filter->nr_excludes = filter->nevents - filter->nr_includes;
+	filter->includes = filter->events;
+	filter->excludes = filter->events + filter->nr_includes;
+
+	return 0;
+}
+
 int kvm_vm_ioctl_set_pmu_event_filter(struct kvm *kvm, void __user *argp)
 {
-	struct kvm_pmu_event_filter tmp, *filter;
+	struct kvm_pmu_event_filter __user *user_filter = argp;
+	struct kvm_x86_pmu_event_filter *filter;
+	struct kvm_pmu_event_filter tmp;
+	struct kvm_vcpu *vcpu;
+	unsigned long i;
 	size_t size;
 	int r;
 
-	if (copy_from_user(&tmp, argp, sizeof(tmp)))
+	if (copy_from_user(&tmp, user_filter, sizeof(tmp)))
 		return -EFAULT;
 
 	if (tmp.action != KVM_PMU_EVENT_ALLOW &&
 	    tmp.action != KVM_PMU_EVENT_DENY)
 		return -EINVAL;
 
-	if (tmp.flags != 0)
+	if (tmp.flags & ~KVM_PMU_EVENT_FLAGS_VALID_MASK)
 		return -EINVAL;
 
 	if (tmp.nevents > KVM_PMU_EVENT_FILTER_MAX_EVENTS)
 		return -E2BIG;
 
 	size = struct_size(filter, events, tmp.nevents);
-	filter = kmalloc(size, GFP_KERNEL_ACCOUNT);
+	filter = kzalloc(size, GFP_KERNEL_ACCOUNT);
 	if (!filter)
 		return -ENOMEM;
 
+	filter->action = tmp.action;
+	filter->nevents = tmp.nevents;
+	filter->fixed_counter_bitmap = tmp.fixed_counter_bitmap;
+	filter->flags = tmp.flags;
+
 	r = -EFAULT;
-	if (copy_from_user(filter, argp, size))
+	if (copy_from_user(filter->events, user_filter->events,
+			   sizeof(filter->events[0]) * filter->nevents))
 		goto cleanup;
 
-	/* Ensure nevents can't be changed between the user copies. */
-	*filter = tmp;
-
-	/*
-	 * Sort the in-kernel list so that we can search it with bsearch.
-	 */
-	sort(&filter->events, filter->nevents, sizeof(__u64), cmp_u64, NULL);
+	r = prepare_filter_lists(filter);
+	if (r)
+		goto cleanup;
 
 	mutex_lock(&kvm->lock);
 	filter = rcu_replace_pointer(kvm->arch.pmu_event_filter, filter,
 				     mutex_is_locked(&kvm->lock));
 	mutex_unlock(&kvm->lock);
-
 	synchronize_srcu_expedited(&kvm->srcu);
+
+	BUILD_BUG_ON(sizeof(((struct kvm_pmu *)0)->reprogram_pmi) >
+		     sizeof(((struct kvm_pmu *)0)->__reprogram_pmi));
+
+	kvm_for_each_vcpu(i, vcpu, kvm)
+		atomic64_set(&vcpu_to_pmu(vcpu)->__reprogram_pmi, -1ull);
+
+	kvm_make_all_cpus_request(kvm, KVM_REQ_PMU);
+
 	r = 0;
 cleanup:
 	kfree(filter);