KVM: MMU: limit rmap chain length

Commit Message

Marcelo Tosatti July 28, 2009, 2:40 a.m. UTC

Otherwise the host can spend too long traversing an rmap chain, which
happens under a spinlock.

Cc: [e-mail]
Signed-off-by: Marcelo Tosatti <[e-mail]>

--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

Avi Kivity Aug. 3, 2009, 12:10 p.m. UTC | #1

On 07/28/2009 05:40 AM, Marcelo Tosatti wrote:
> Otherwise the host can spend too long traversing an rmap chain, which
> happens under a spinlock.
>
>    

Applied, thanks.

Patch

Index: kvm/arch/x86/kvm/mmu.c
===================================================================
--- kvm.orig/arch/x86/kvm/mmu.c
+++ kvm/arch/x86/kvm/mmu.c
@@ -502,16 +502,20 @@  static unsigned long *gfn_to_rmap(struct
  *
  * If rmapp bit zero is one, (then rmap & ~1) points to a struct kvm_rmap_desc
  * containing more mappings.
+ *
+ * Returns the number of rmap entries before the spte was added or zero if
+ * the spte was not added.
+ *
  */
-static void rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn, int lpage)
+static int rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn, int lpage)
 {
 	struct kvm_mmu_page *sp;
 	struct kvm_rmap_desc *desc;
 	unsigned long *rmapp;
-	int i;
+	int i, count = 0;
 
 	if (!is_rmap_spte(*spte))
-		return;
+		return count;
 	gfn = unalias_gfn(vcpu->kvm, gfn);
 	sp = page_header(__pa(spte));
 	sp->gfns[spte - sp->spt] = gfn;
@@ -528,8 +532,10 @@  static void rmap_add(struct kvm_vcpu *vc
 	} else {
 		rmap_printk("rmap_add: %p %llx many->many\n", spte, *spte);
 		desc = (struct kvm_rmap_desc *)(*rmapp & ~1ul);
-		while (desc->sptes[RMAP_EXT-1] && desc->more)
+		while (desc->sptes[RMAP_EXT-1] && desc->more) {
 			desc = desc->more;
+			count += RMAP_EXT;
+		}
 		if (desc->sptes[RMAP_EXT-1]) {
 			desc->more = mmu_alloc_rmap_desc(vcpu);
 			desc = desc->more;
@@ -538,6 +544,7 @@  static void rmap_add(struct kvm_vcpu *vc
 			;
 		desc->sptes[i] = spte;
 	}
+	return count;
 }
 
 static void rmap_desc_remove_entry(unsigned long *rmapp,
@@ -767,6 +774,19 @@  static int kvm_age_rmapp(struct kvm *kvm
 	return young;
 }
 
+#define RMAP_RECYCLE_THRESHOLD 1000
+
+static void rmap_recycle(struct kvm_vcpu *vcpu, gfn_t gfn, int lpage)
+{
+	unsigned long *rmapp;
+
+	gfn = unalias_gfn(vcpu->kvm, gfn);
+	rmapp = gfn_to_rmap(vcpu->kvm, gfn, lpage);
+
+	kvm_unmap_rmapp(vcpu->kvm, rmapp);
+	kvm_flush_remote_tlbs(vcpu->kvm);
+}
+
 int kvm_age_hva(struct kvm *kvm, unsigned long hva)
 {
 	return kvm_handle_hva(kvm, hva, kvm_age_rmapp);
@@ -1753,6 +1773,7 @@  static void mmu_set_spte(struct kvm_vcpu
 {
 	int was_rmapped = 0;
 	int was_writeble = is_writeble_pte(*sptep);
+	int rmap_count;
 
 	pgprintk("%s: spte %llx access %x write_fault %d"
 		 " user_fault %d gfn %lx\n",
@@ -1794,9 +1815,11 @@  static void mmu_set_spte(struct kvm_vcpu
 
 	page_header_update_slot(vcpu->kvm, sptep, gfn);
 	if (!was_rmapped) {
-		rmap_add(vcpu, sptep, gfn, largepage);
+		rmap_count = rmap_add(vcpu, sptep, gfn, largepage);
 		if (!is_rmap_spte(*sptep))
 			kvm_release_pfn_clean(pfn);
+		if (rmap_count > RMAP_RECYCLE_THRESHOLD)
+			rmap_recycle(vcpu, gfn, largepage);
 	} else {
 		if (was_writeble)
 			kvm_release_pfn_dirty(pfn);