1 files changed, 366 insertions, 94 deletions

diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 676edfc19a95..3a281a2decde 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -37,6 +37,7 @@
 #include <linux/srcu.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
+#include <linux/kthread.h>
 
 #include <asm/page.h>
 #include <asm/cmpxchg.h>
@@ -44,6 +45,30 @@
 #include <asm/vmx.h>
 #include <asm/kvm_page_track.h>
 
+extern bool itlb_multihit_kvm_mitigation;
+
+static int __read_mostly nx_huge_pages = -1;
+static uint __read_mostly nx_huge_pages_recovery_ratio = 60;
+
+static int set_nx_huge_pages(const char *val, const struct kernel_param *kp);
+static int set_nx_huge_pages_recovery_ratio(const char *val, const struct kernel_param *kp);
+
+static struct kernel_param_ops nx_huge_pages_ops = {
+	.set = set_nx_huge_pages,
+	.get = param_get_bool,
+};
+
+static struct kernel_param_ops nx_huge_pages_recovery_ratio_ops = {
+	.set = set_nx_huge_pages_recovery_ratio,
+	.get = param_get_uint,
+};
+
+module_param_cb(nx_huge_pages, &nx_huge_pages_ops, &nx_huge_pages, 0644);
+__MODULE_PARM_TYPE(nx_huge_pages, "bool");
+module_param_cb(nx_huge_pages_recovery_ratio, &nx_huge_pages_recovery_ratio_ops,
+		&nx_huge_pages_recovery_ratio, 0644);
+__MODULE_PARM_TYPE(nx_huge_pages_recovery_ratio, "uint");
+
 /*
  * When setting this variable to true it enables Two-Dimensional-Paging
  * where the hardware walks 2 page tables:
@@ -131,9 +156,6 @@ module_param(dbg, bool, 0644);
 
 #include <trace/events/kvm.h>
 
-#define CREATE_TRACE_POINTS
-#include "mmutrace.h"
-
 #define SPTE_HOST_WRITEABLE	(1ULL << PT_FIRST_AVAIL_BITS_SHIFT)
 #define SPTE_MMU_WRITEABLE	(1ULL << (PT_FIRST_AVAIL_BITS_SHIFT + 1))
 
@@ -142,6 +164,20 @@ module_param(dbg, bool, 0644);
 /* make pte_list_desc fit well in cache line */
 #define PTE_LIST_EXT 3
 
+/*
+ * Return values of handle_mmio_page_fault and mmu.page_fault:
+ * RET_PF_RETRY: let CPU fault again on the address.
+ * RET_PF_EMULATE: mmio page fault, emulate the instruction directly.
+ *
+ * For handle_mmio_page_fault only:
+ * RET_PF_INVALID: the spte is invalid, let the real page fault path update it.
+ */
+enum {
+	RET_PF_RETRY = 0,
+	RET_PF_EMULATE = 1,
+	RET_PF_INVALID = 2,
+};
+
 struct pte_list_desc {
 	u64 *sptes[PTE_LIST_EXT];
 	struct pte_list_desc *more;
@@ -179,14 +215,23 @@ static u64 __read_mostly shadow_mmio_mask;
 static u64 __read_mostly shadow_present_mask;
 
 static void mmu_spte_set(u64 *sptep, u64 spte);
+static bool is_executable_pte(u64 spte);
 static void mmu_free_roots(struct kvm_vcpu *vcpu);
 
+#define CREATE_TRACE_POINTS
+#include "mmutrace.h"
+
 void kvm_mmu_set_mmio_spte_mask(u64 mmio_mask)
 {
 	shadow_mmio_mask = mmio_mask;
 }
 EXPORT_SYMBOL_GPL(kvm_mmu_set_mmio_spte_mask);
 
+static bool is_nx_huge_page_enabled(void)
+{
+	return READ_ONCE(nx_huge_pages);
+}
+
 /*
  * the low bit of the generation number is always presumed to be zero.
  * This disables mmio caching during memslot updates.  The concept is
@@ -324,6 +369,11 @@ static int is_last_spte(u64 pte, int level)
 	return 0;
 }
 
+static bool is_executable_pte(u64 spte)
+{
+	return (spte & (shadow_x_mask | shadow_nx_mask)) == shadow_x_mask;
+}
+
 static kvm_pfn_t spte_to_pfn(u64 pte)
 {
 	return (pte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
@@ -767,10 +817,16 @@ static gfn_t kvm_mmu_page_get_gfn(struct kvm_mmu_page *sp, int index)
 
 static void kvm_mmu_page_set_gfn(struct kvm_mmu_page *sp, int index, gfn_t gfn)
 {
-	if (sp->role.direct)
-		BUG_ON(gfn != kvm_mmu_page_get_gfn(sp, index));
-	else
+	if (!sp->role.direct) {
 		sp->gfns[index] = gfn;
+		return;
+	}
+
+	if (WARN_ON(gfn != kvm_mmu_page_get_gfn(sp, index)))
+		pr_err_ratelimited("gfn mismatch under direct page %llx "
+				   "(expected %llx, got %llx)\n",
+				   sp->gfn,
+				   kvm_mmu_page_get_gfn(sp, index), gfn);
 }
 
 /*
@@ -829,6 +885,17 @@ static void account_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
 	kvm_mmu_gfn_disallow_lpage(slot, gfn);
 }
 
+static void account_huge_nx_page(struct kvm *kvm, struct kvm_mmu_page *sp)
+{
+	if (sp->lpage_disallowed)
+		return;
+
+	++kvm->stat.nx_lpage_splits;
+	list_add_tail(&sp->lpage_disallowed_link,
+		      &kvm->arch.lpage_disallowed_mmu_pages);
+	sp->lpage_disallowed = true;
+}
+
 static void unaccount_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
 {
 	struct kvm_memslots *slots;
@@ -846,6 +913,13 @@ static void unaccount_shadowed(struct kvm *kvm, struct kvm_mmu_page *sp)
 	kvm_mmu_gfn_allow_lpage(slot, gfn);
 }
 
+static void unaccount_huge_nx_page(struct kvm *kvm, struct kvm_mmu_page *sp)
+{
+	--kvm->stat.nx_lpage_splits;
+	sp->lpage_disallowed = false;
+	list_del(&sp->lpage_disallowed_link);
+}
+
 static bool __mmu_gfn_lpage_is_disallowed(gfn_t gfn, int level,
 					  struct kvm_memory_slot *slot)
 {
@@ -2382,6 +2456,9 @@ static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
 			kvm_reload_remote_mmus(kvm);
 	}
 
+	if (sp->lpage_disallowed)
+		unaccount_huge_nx_page(kvm, sp);
+
 	sp->role.invalid = 1;
 	return ret;
 }
@@ -2533,6 +2610,11 @@ static int set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
 	if (!speculative)
 		spte |= shadow_accessed_mask;
 
+	if (level > PT_PAGE_TABLE_LEVEL && (pte_access & ACC_EXEC_MASK) &&
+	    is_nx_huge_page_enabled()) {
+		pte_access &= ~ACC_EXEC_MASK;
+	}
+
 	if (pte_access & ACC_EXEC_MASK)
 		spte |= shadow_x_mask;
 	else
@@ -2598,13 +2680,13 @@ done:
 	return ret;
 }
 
-static bool mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep, unsigned pte_access,
-			 int write_fault, int level, gfn_t gfn, kvm_pfn_t pfn,
-			 bool speculative, bool host_writable)
+static int mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep, unsigned pte_access,
+			int write_fault, int level, gfn_t gfn, kvm_pfn_t pfn,
+		       	bool speculative, bool host_writable)
 {
 	int was_rmapped = 0;
 	int rmap_count;
-	bool emulate = false;
+	int ret = RET_PF_RETRY;
 
 	pgprintk("%s: spte %llx write_fault %d gfn %llx\n", __func__,
 		 *sptep, write_fault, gfn);
@@ -2634,18 +2716,15 @@ static bool mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep, unsigned pte_access,
 	if (set_spte(vcpu, sptep, pte_access, level, gfn, pfn, speculative,
 	      true, host_writable)) {
 		if (write_fault)
-			emulate = true;
+			ret = RET_PF_EMULATE;
 		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
 	}
 
 	if (unlikely(is_mmio_spte(*sptep)))
-		emulate = true;
+		ret = RET_PF_EMULATE;
 
 	pgprintk("%s: setting spte %llx\n", __func__, *sptep);
-	pgprintk("instantiating %s PTE (%s) at %llx (%llx) addr %p\n",
-		 is_large_pte(*sptep)? "2MB" : "4kB",
-		 *sptep & PT_PRESENT_MASK ?"RW":"R", gfn,
-		 *sptep, sptep);
+	trace_kvm_mmu_set_spte(level, gfn, sptep);
 	if (!was_rmapped && is_large_pte(*sptep))
 		++vcpu->kvm->stat.lpages;
 
@@ -2657,9 +2736,7 @@ static bool mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep, unsigned pte_access,
 		}
 	}
 
-	kvm_release_pfn_clean(pfn);
-
-	return emulate;
+	return ret;
 }
 
 static kvm_pfn_t pte_prefetch_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn,
@@ -2693,9 +2770,11 @@ static int direct_pte_prefetch_many(struct kvm_vcpu *vcpu,
 	if (ret <= 0)
 		return -1;
 
-	for (i = 0; i < ret; i++, gfn++, start++)
+	for (i = 0; i < ret; i++, gfn++, start++) {
 		mmu_set_spte(vcpu, start, access, 0, sp->role.level, gfn,
 			     page_to_pfn(pages[i]), true, true);
+		put_page(pages[i]);
+	}
 
 	return 0;
 }
@@ -2743,40 +2822,71 @@ static void direct_pte_prefetch(struct kvm_vcpu *vcpu, u64 *sptep)
 	__direct_pte_prefetch(vcpu, sp, sptep);
 }
 
-static int __direct_map(struct kvm_vcpu *vcpu, int write, int map_writable,
-			int level, gfn_t gfn, kvm_pfn_t pfn, bool prefault)
+static void disallowed_hugepage_adjust(struct kvm_shadow_walk_iterator it,
+				       gfn_t gfn, kvm_pfn_t *pfnp, int *levelp)
 {
-	struct kvm_shadow_walk_iterator iterator;
+	int level = *levelp;
+	u64 spte = *it.sptep;
+
+	if (it.level == level && level > PT_PAGE_TABLE_LEVEL &&
+	    is_nx_huge_page_enabled() &&
+	    is_shadow_present_pte(spte) &&
+	    !is_large_pte(spte)) {
+		/*
+		 * A small SPTE exists for this pfn, but FNAME(fetch)
+		 * and __direct_map would like to create a large PTE
+		 * instead: just force them to go down another level,
+		 * patching back for them into pfn the next 9 bits of
+		 * the address.
+		 */
+		u64 page_mask = KVM_PAGES_PER_HPAGE(level) - KVM_PAGES_PER_HPAGE(level - 1);
+		*pfnp |= gfn & page_mask;
+		(*levelp)--;
+	}
+}
+
+static int __direct_map(struct kvm_vcpu *vcpu, gpa_t gpa, int write,
+			int map_writable, int level, kvm_pfn_t pfn,
+			bool prefault, bool lpage_disallowed)
+{
+	struct kvm_shadow_walk_iterator it;
 	struct kvm_mmu_page *sp;
-	int emulate = 0;
-	gfn_t pseudo_gfn;
+	int ret;
+	gfn_t gfn = gpa >> PAGE_SHIFT;
+	gfn_t base_gfn = gfn;
 
 	if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
-		return 0;
+		return RET_PF_RETRY;
 
-	for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
-		if (iterator.level == level) {
-			emulate = mmu_set_spte(vcpu, iterator.sptep, ACC_ALL,
-					       write, level, gfn, pfn, prefault,
-					       map_writable);
-			direct_pte_prefetch(vcpu, iterator.sptep);
-			++vcpu->stat.pf_fixed;
-			break;
-		}
+	trace_kvm_mmu_spte_requested(gpa, level, pfn);
+	for_each_shadow_entry(vcpu, gpa, it) {
+		/*
+		 * We cannot overwrite existing page tables with an NX
+		 * large page, as the leaf could be executable.
+		 */
+		disallowed_hugepage_adjust(it, gfn, &pfn, &level);
 
-		drop_large_spte(vcpu, iterator.sptep);
-		if (!is_shadow_present_pte(*iterator.sptep)) {
-			u64 base_addr = iterator.addr;
+		base_gfn = gfn & ~(KVM_PAGES_PER_HPAGE(it.level) - 1);
+		if (it.level == level)
+			break;
 
-			base_addr &= PT64_LVL_ADDR_MASK(iterator.level);
-			pseudo_gfn = base_addr >> PAGE_SHIFT;
-			sp = kvm_mmu_get_page(vcpu, pseudo_gfn, iterator.addr,
-					      iterator.level - 1, 1, ACC_ALL);
+		drop_large_spte(vcpu, it.sptep);
+		if (!is_shadow_present_pte(*it.sptep)) {
+			sp = kvm_mmu_get_page(vcpu, base_gfn, it.addr,
+					      it.level - 1, true, ACC_ALL);
 
-			link_shadow_page(vcpu, iterator.sptep, sp);
+			link_shadow_page(vcpu, it.sptep, sp);
+			if (lpage_disallowed)
+				account_huge_nx_page(vcpu->kvm, sp);
 		}
 	}
-	return emulate;
+
+	ret = mmu_set_spte(vcpu, it.sptep, ACC_ALL,
+			   write, level, base_gfn, pfn, prefault,
+			   map_writable);
+	direct_pte_prefetch(vcpu, it.sptep);
+	++vcpu->stat.pf_fixed;
+	return ret;
 }
 
 static void kvm_send_hwpoison_signal(unsigned long address, struct task_struct *tsk)
@@ -2798,25 +2908,23 @@ static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, kvm_pfn_t pfn)
 	 * Do not cache the mmio info caused by writing the readonly gfn
 	 * into the spte otherwise read access on readonly gfn also can
 	 * caused mmio page fault and treat it as mmio access.
-	 * Return 1 to tell kvm to emulate it.
 	 */
 	if (pfn == KVM_PFN_ERR_RO_FAULT)
-		return 1;
+		return RET_PF_EMULATE;
 
 	if (pfn == KVM_PFN_ERR_HWPOISON) {
 		kvm_send_hwpoison_signal(kvm_vcpu_gfn_to_hva(vcpu, gfn), current);
-		return 0;
+		return RET_PF_RETRY;
 	}
 
 	return -EFAULT;
 }
 
 static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
-					gfn_t *gfnp, kvm_pfn_t *pfnp,
+					gfn_t gfn, kvm_pfn_t *pfnp,
 					int *levelp)
 {
 	kvm_pfn_t pfn = *pfnp;
-	gfn_t gfn = *gfnp;
 	int level = *levelp;
 
 	/*
@@ -2826,7 +2934,7 @@ static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 	 * here.
 	 */
 	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
-	    level == PT_PAGE_TABLE_LEVEL &&
+	    !kvm_is_zone_device_pfn(pfn) && level == PT_PAGE_TABLE_LEVEL &&
 	    PageTransCompoundMap(pfn_to_page(pfn)) &&
 	    !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
 		unsigned long mask;
@@ -2843,8 +2951,6 @@ static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 		mask = KVM_PAGES_PER_HPAGE(level) - 1;
 		VM_BUG_ON((gfn & mask) != (pfn & mask));
 		if (pfn & mask) {
-			gfn &= ~mask;
-			*gfnp = gfn;
 			kvm_release_pfn_clean(pfn);
 			pfn &= ~mask;
 			kvm_get_pfn(pfn);
@@ -3012,11 +3118,14 @@ static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code,
 {
 	int r;
 	int level;
-	bool force_pt_level = false;
+	bool force_pt_level;
 	kvm_pfn_t pfn;
 	unsigned long mmu_seq;
 	bool map_writable, write = error_code & PFERR_WRITE_MASK;
+	bool lpage_disallowed = (error_code & PFERR_FETCH_MASK) &&
+				is_nx_huge_page_enabled();
 
+	force_pt_level = lpage_disallowed;
 	level = mapping_level(vcpu, gfn, &force_pt_level);
 	if (likely(!force_pt_level)) {
 		/*
@@ -3031,32 +3140,30 @@ static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code,
 	}
 
 	if (fast_page_fault(vcpu, v, level, error_code))
-		return 0;
+		return RET_PF_RETRY;
 
 	mmu_seq = vcpu->kvm->mmu_notifier_seq;
 	smp_rmb();
 
 	if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable))
-		return 0;
+		return RET_PF_RETRY;
 
 	if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r))
 		return r;
 
+	r = RET_PF_RETRY;
 	spin_lock(&vcpu->kvm->mmu_lock);
 	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
 		goto out_unlock;
 	make_mmu_pages_available(vcpu);
 	if (likely(!force_pt_level))
-		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
-	r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault);
-	spin_unlock(&vcpu->kvm->mmu_lock);
-
-	return r;
-
+		transparent_hugepage_adjust(vcpu, gfn, &pfn, &level);
+	r = __direct_map(vcpu, v, write, map_writable, level, pfn,
+			 prefault, false);
 out_unlock:
 	spin_unlock(&vcpu->kvm->mmu_lock);
 	kvm_release_pfn_clean(pfn);
-	return 0;
+	return r;
 }
 
 
@@ -3383,38 +3490,38 @@ exit:
 	return reserved;
 }
 
-int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr, bool direct)
+static int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr, bool direct)
 {
 	u64 spte;
 	bool reserved;
 
 	if (mmio_info_in_cache(vcpu, addr, direct))
-		return RET_MMIO_PF_EMULATE;
+		return RET_PF_EMULATE;
 
 	reserved = walk_shadow_page_get_mmio_spte(vcpu, addr, &spte);
 	if (WARN_ON(reserved))
-		return RET_MMIO_PF_BUG;
+		return -EINVAL;
 
 	if (is_mmio_spte(spte)) {
 		gfn_t gfn = get_mmio_spte_gfn(spte);
 		unsigned access = get_mmio_spte_access(spte);
 
 		if (!check_mmio_spte(vcpu, spte))
-			return RET_MMIO_PF_INVALID;
+			return RET_PF_INVALID;
 
 		if (direct)
 			addr = 0;
 
 		trace_handle_mmio_page_fault(addr, gfn, access);
 		vcpu_cache_mmio_info(vcpu, addr, gfn, access);
-		return RET_MMIO_PF_EMULATE;
+		return RET_PF_EMULATE;
 	}
 
 	/*
 	 * If the page table is zapped by other cpus, let CPU fault again on
 	 * the address.
 	 */
-	return RET_MMIO_PF_RETRY;
+	return RET_PF_RETRY;
 }
 EXPORT_SYMBOL_GPL(handle_mmio_page_fault);
 
@@ -3464,7 +3571,7 @@ static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
 	pgprintk("%s: gva %lx error %x\n", __func__, gva, error_code);
 
 	if (page_fault_handle_page_track(vcpu, error_code, gfn))
-		return 1;
+		return RET_PF_EMULATE;
 
 	r = mmu_topup_memory_caches(vcpu);
 	if (r)
@@ -3548,18 +3655,21 @@ static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
 	unsigned long mmu_seq;
 	int write = error_code & PFERR_WRITE_MASK;
 	bool map_writable;
+	bool lpage_disallowed = (error_code & PFERR_FETCH_MASK) &&
+				is_nx_huge_page_enabled();
 
 	MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa));
 
 	if (page_fault_handle_page_track(vcpu, error_code, gfn))
-		return 1;
+		return RET_PF_EMULATE;
 
 	r = mmu_topup_memory_caches(vcpu);
 	if (r)
 		return r;
 
-	force_pt_level = !check_hugepage_cache_consistency(vcpu, gfn,
-							   PT_DIRECTORY_LEVEL);
+	force_pt_level =
+		lpage_disallowed ||
+		!check_hugepage_cache_consistency(vcpu, gfn, PT_DIRECTORY_LEVEL);
 	level = mapping_level(vcpu, gfn, &force_pt_level);
 	if (likely(!force_pt_level)) {
 		if (level > PT_DIRECTORY_LEVEL &&
@@ -3569,32 +3679,30 @@ static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
 	}
 
 	if (fast_page_fault(vcpu, gpa, level, error_code))
-		return 0;
+		return RET_PF_RETRY;
 
 	mmu_seq = vcpu->kvm->mmu_notifier_seq;
 	smp_rmb();
 
 	if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable))
-		return 0;
+		return RET_PF_RETRY;
 
 	if (handle_abnormal_pfn(vcpu, 0, gfn, pfn, ACC_ALL, &r))
 		return r;
 
+	r = RET_PF_RETRY;
 	spin_lock(&vcpu->kvm->mmu_lock);
 	if (mmu_notifier_retry(vcpu->kvm, mmu_seq))
 		goto out_unlock;
 	make_mmu_pages_available(vcpu);
 	if (likely(!force_pt_level))
-		transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level);
-	r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault);
-	spin_unlock(&vcpu->kvm->mmu_lock);
-
-	return r;
-
+		transparent_hugepage_adjust(vcpu, gfn, &pfn, &level);
+	r = __direct_map(vcpu, gpa, write, map_writable, level, pfn,
+			 prefault, lpage_disallowed);
 out_unlock:
 	spin_unlock(&vcpu->kvm->mmu_lock);
 	kvm_release_pfn_clean(pfn);
-	return 0;
+	return r;
 }
 
 static void nonpaging_init_context(struct kvm_vcpu *vcpu,
@@ -4510,23 +4618,24 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u32 error_code,
 	enum emulation_result er;
 	bool direct = vcpu->arch.mmu.direct_map || mmu_is_nested(vcpu);
 
+	r = RET_PF_INVALID;
 	if (unlikely(error_code & PFERR_RSVD_MASK)) {
 		r = handle_mmio_page_fault(vcpu, cr2, direct);
-		if (r == RET_MMIO_PF_EMULATE) {
+		if (r == RET_PF_EMULATE) {
 			emulation_type = 0;
 			goto emulate;
 		}
-		if (r == RET_MMIO_PF_RETRY)
-			return 1;
-		if (r < 0)
-			return r;
 	}
 
-	r = vcpu->arch.mmu.page_fault(vcpu, cr2, error_code, false);
+	if (r == RET_PF_INVALID) {
+		r = vcpu->arch.mmu.page_fault(vcpu, cr2, error_code, false);
+		WARN_ON(r == RET_PF_INVALID);
+	}
+
+	if (r == RET_PF_RETRY)
+		return 1;
 	if (r < 0)
 		return r;
-	if (!r)
-		return 1;
 
 	if (mmio_info_in_cache(vcpu, cr2, direct))
 		emulation_type = 0;
@@ -4781,9 +4890,9 @@ restart:
 		 * the guest, and the guest page table is using 4K page size
 		 * mapping if the indirect sp has level = 1.
 		 */
-		if (sp->role.direct &&
-			!kvm_is_reserved_pfn(pfn) &&
-			PageTransCompoundMap(pfn_to_page(pfn))) {
+		if (sp->role.direct && !kvm_is_reserved_pfn(pfn) &&
+		    !kvm_is_zone_device_pfn(pfn) &&
+		    PageTransCompoundMap(pfn_to_page(pfn))) {
 			drop_spte(kvm, sptep);
 			need_tlb_flush = 1;
 			goto restart;
@@ -4965,7 +5074,7 @@ mmu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
 	int nr_to_scan = sc->nr_to_scan;
 	unsigned long freed = 0;
 
-	spin_lock(&kvm_lock);
+	mutex_lock(&kvm_lock);
 
 	list_for_each_entry(kvm, &vm_list, vm_list) {
 		int idx;
@@ -5015,7 +5124,7 @@ unlock:
 		break;
 	}
 
-	spin_unlock(&kvm_lock);
+	mutex_unlock(&kvm_lock);
 	return freed;
 }
 
@@ -5039,8 +5148,58 @@ static void mmu_destroy_caches(void)
 		kmem_cache_destroy(mmu_page_header_cache);
 }
 
+static bool get_nx_auto_mode(void)
+{
+	/* Return true when CPU has the bug, and mitigations are ON */
+	return boot_cpu_has_bug(X86_BUG_ITLB_MULTIHIT) && !cpu_mitigations_off();
+}
+
+static void __set_nx_huge_pages(bool val)
+{
+	nx_huge_pages = itlb_multihit_kvm_mitigation = val;
+}
+
+static int set_nx_huge_pages(const char *val, const struct kernel_param *kp)
+{
+	bool old_val = nx_huge_pages;
+	bool new_val;
+
+	/* In "auto" mode deploy workaround only if CPU has the bug. */
+	if (sysfs_streq(val, "off"))
+		new_val = 0;
+	else if (sysfs_streq(val, "force"))
+		new_val = 1;
+	else if (sysfs_streq(val, "auto"))
+		new_val = get_nx_auto_mode();
+	else if (strtobool(val, &new_val) < 0)
+		return -EINVAL;
+
+	__set_nx_huge_pages(new_val);
+
+	if (new_val != old_val) {
+		struct kvm *kvm;
+		int idx;
+
+		mutex_lock(&kvm_lock);
+
+		list_for_each_entry(kvm, &vm_list, vm_list) {
+			idx = srcu_read_lock(&kvm->srcu);
+			kvm_mmu_invalidate_zap_all_pages(kvm);
+			srcu_read_unlock(&kvm->srcu, idx);
+
+			wake_up_process(kvm->arch.nx_lpage_recovery_thread);
+		}
+		mutex_unlock(&kvm_lock);
+	}
+
+	return 0;
+}
+
 int kvm_mmu_module_init(void)
 {
+	if (nx_huge_pages == -1)
+		__set_nx_huge_pages(get_nx_auto_mode());
+
 	pte_list_desc_cache = kmem_cache_create("pte_list_desc",
 					    sizeof(struct pte_list_desc),
 					    0, SLAB_ACCOUNT, NULL);
@@ -5104,3 +5263,116 @@ void kvm_mmu_module_exit(void)
 	unregister_shrinker(&mmu_shrinker);
 	mmu_audit_disable();
 }
+
+static int set_nx_huge_pages_recovery_ratio(const char *val, const struct kernel_param *kp)
+{
+	unsigned int old_val;
+	int err;
+
+	old_val = nx_huge_pages_recovery_ratio;
+	err = param_set_uint(val, kp);
+	if (err)
+		return err;
+
+	if (READ_ONCE(nx_huge_pages) &&
+	    !old_val && nx_huge_pages_recovery_ratio) {
+		struct kvm *kvm;
+
+		mutex_lock(&kvm_lock);
+
+		list_for_each_entry(kvm, &vm_list, vm_list)
+			wake_up_process(kvm->arch.nx_lpage_recovery_thread);
+
+		mutex_unlock(&kvm_lock);
+	}
+
+	return err;
+}
+
+static void kvm_recover_nx_lpages(struct kvm *kvm)
+{
+	int rcu_idx;
+	struct kvm_mmu_page *sp;
+	unsigned int ratio;
+	LIST_HEAD(invalid_list);
+	ulong to_zap;
+
+	rcu_idx = srcu_read_lock(&kvm->srcu);
+	spin_lock(&kvm->mmu_lock);
+
+	ratio = READ_ONCE(nx_huge_pages_recovery_ratio);
+	to_zap = ratio ? DIV_ROUND_UP(kvm->stat.nx_lpage_splits, ratio) : 0;
+	while (to_zap && !list_empty(&kvm->arch.lpage_disallowed_mmu_pages)) {
+		/*
+		 * We use a separate list instead of just using active_mmu_pages
+		 * because the number of lpage_disallowed pages is expected to
+		 * be relatively small compared to the total.
+		 */
+		sp = list_first_entry(&kvm->arch.lpage_disallowed_mmu_pages,
+				      struct kvm_mmu_page,
+				      lpage_disallowed_link);
+		WARN_ON_ONCE(!sp->lpage_disallowed);
+		kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list);
+		WARN_ON_ONCE(sp->lpage_disallowed);
+
+		if (!--to_zap || need_resched() || spin_needbreak(&kvm->mmu_lock)) {
+			kvm_mmu_commit_zap_page(kvm, &invalid_list);
+			if (to_zap)
+				cond_resched_lock(&kvm->mmu_lock);
+		}
+	}
+
+	spin_unlock(&kvm->mmu_lock);
+	srcu_read_unlock(&kvm->srcu, rcu_idx);
+}
+
+static long get_nx_lpage_recovery_timeout(u64 start_time)
+{
+	return READ_ONCE(nx_huge_pages) && READ_ONCE(nx_huge_pages_recovery_ratio)
+		? start_time + 60 * HZ - get_jiffies_64()
+		: MAX_SCHEDULE_TIMEOUT;
+}
+
+static int kvm_nx_lpage_recovery_worker(struct kvm *kvm, uintptr_t data)
+{
+	u64 start_time;
+	long remaining_time;
+
+	while (true) {
+		start_time = get_jiffies_64();
+		remaining_time = get_nx_lpage_recovery_timeout(start_time);
+
+		set_current_state(TASK_INTERRUPTIBLE);
+		while (!kthread_should_stop() && remaining_time > 0) {
+			schedule_timeout(remaining_time);
+			remaining_time = get_nx_lpage_recovery_timeout(start_time);
+			set_current_state(TASK_INTERRUPTIBLE);
+		}
+
+		set_current_state(TASK_RUNNING);
+
+		if (kthread_should_stop())
+			return 0;
+
+		kvm_recover_nx_lpages(kvm);
+	}
+}
+
+int kvm_mmu_post_init_vm(struct kvm *kvm)
+{
+	int err;
+
+	err = kvm_vm_create_worker_thread(kvm, kvm_nx_lpage_recovery_worker, 0,
+					  "kvm-nx-lpage-recovery",
+					  &kvm->arch.nx_lpage_recovery_thread);
+	if (!err)
+		kthread_unpark(kvm->arch.nx_lpage_recovery_thread);
+
+	return err;
+}
+
+void kvm_mmu_pre_destroy_vm(struct kvm *kvm)
+{
+	if (kvm->arch.nx_lpage_recovery_thread)
+		kthread_stop(kvm->arch.nx_lpage_recovery_thread);
+}