KVM: Avoid saving and restoring some host CPU state on lightweight vmexit

Many msrs and the like will only be used by the host if we schedule() or
return to userspace.  Therefore, we avoid saving them if we handle the
exit within the kernel, and if a reschedule is not requested.

Based on a patch from Eddie Dong <eddie.dong@intel.com> with a couple of
fixes by me.

Signed-off-by: Yaozu(Eddie) Dong <eddie.dong@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>

3 files changed, 62 insertions, 45 deletions

diff --git a/drivers/kvm/kvm.h b/drivers/kvm/kvm.h
index 152312c1fafa..7facebd1911d 100644
--- a/drivers/kvm/kvm.h
+++ b/drivers/kvm/kvm.h
@@ -252,6 +252,7 @@ struct kvm_stat {
 	u32 halt_exits;
 	u32 request_irq_exits;
 	u32 irq_exits;
+	u32 light_exits;
 };
 
 struct kvm_vcpu {
diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index 8f1f07adb04e..7d682586423b 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -72,6 +72,7 @@ static struct kvm_stats_debugfs_item {
 	{ "halt_exits", STAT_OFFSET(halt_exits) },
 	{ "request_irq", STAT_OFFSET(request_irq_exits) },
 	{ "irq_exits", STAT_OFFSET(irq_exits) },
+	{ "light_exits", STAT_OFFSET(light_exits) },
 	{ NULL }
 };
 
diff --git a/drivers/kvm/vmx.c b/drivers/kvm/vmx.c
index 52bd5f079df1..84ce0c0930a0 100644
--- a/drivers/kvm/vmx.c
+++ b/drivers/kvm/vmx.c
@@ -483,6 +483,13 @@ static int vmx_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data)
 	case MSR_GS_BASE:
 		vmcs_writel(GUEST_GS_BASE, data);
 		break;
+	case MSR_LSTAR:
+	case MSR_SYSCALL_MASK:
+		msr = find_msr_entry(vcpu, msr_index);
+		if (msr)
+			msr->data = data;
+		load_msrs(vcpu->guest_msrs, NR_BAD_MSRS);
+		break;
 #endif
 	case MSR_IA32_SYSENTER_CS:
 		vmcs_write32(GUEST_SYSENTER_CS, data);
@@ -1820,7 +1827,7 @@ static int vmx_vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 	int fs_gs_ldt_reload_needed;
 	int r;
 
-again:
+preempted:
 	/*
 	 * Set host fs and gs selectors.  Unfortunately, 22.2.3 does not
 	 * allow segment selectors with cpl > 0 or ti == 1.
@@ -1851,13 +1858,6 @@ again:
 	if (vcpu->guest_debug.enabled)
 		kvm_guest_debug_pre(vcpu);
 
-	kvm_load_guest_fpu(vcpu);
-
-	/*
-	 * Loading guest fpu may have cleared host cr0.ts
-	 */
-	vmcs_writel(HOST_CR0, read_cr0());
-
 #ifdef CONFIG_X86_64
 	if (is_long_mode(vcpu)) {
 		save_msrs(vcpu->host_msrs + msr_offset_kernel_gs_base, 1);
@@ -1865,6 +1865,14 @@ again:
 	}
 #endif
 
+again:
+	kvm_load_guest_fpu(vcpu);
+
+	/*
+	 * Loading guest fpu may have cleared host cr0.ts
+	 */
+	vmcs_writel(HOST_CR0, read_cr0());
+
 	asm (
 		/* Store host registers */
 		"pushf \n\t"
@@ -1984,36 +1992,8 @@ again:
 		[cr2]"i"(offsetof(struct kvm_vcpu, cr2))
 	      : "cc", "memory" );
 
-	/*
-	 * Reload segment selectors ASAP. (it's needed for a functional
-	 * kernel: x86 relies on having __KERNEL_PDA in %fs and x86_64
-	 * relies on having 0 in %gs for the CPU PDA to work.)
-	 */
-	if (fs_gs_ldt_reload_needed) {
-		load_ldt(ldt_sel);
-		load_fs(fs_sel);
-		/*
-		 * If we have to reload gs, we must take care to
-		 * preserve our gs base.
-		 */
-		local_irq_disable();
-		load_gs(gs_sel);
-#ifdef CONFIG_X86_64
-		wrmsrl(MSR_GS_BASE, vmcs_readl(HOST_GS_BASE));
-#endif
-		local_irq_enable();
-
-		reload_tss();
-	}
 	++vcpu->stat.exits;
 
-#ifdef CONFIG_X86_64
-	if (is_long_mode(vcpu)) {
-		save_msrs(vcpu->guest_msrs, NR_BAD_MSRS);
-		load_msrs(vcpu->host_msrs, NR_BAD_MSRS);
-	}
-#endif
-
 	vcpu->interrupt_window_open = (vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & 3) == 0;
 
 	asm ("mov %0, %%ds; mov %0, %%es" : : "r"(__USER_DS));
@@ -2035,24 +2015,59 @@ again:
 		if (r > 0) {
 			/* Give scheduler a change to reschedule. */
 			if (signal_pending(current)) {
-				++vcpu->stat.signal_exits;
-				post_kvm_run_save(vcpu, kvm_run);
+				r = -EINTR;
 				kvm_run->exit_reason = KVM_EXIT_INTR;
-				return -EINTR;
+				++vcpu->stat.signal_exits;
+				goto out;
 			}
 
 			if (dm_request_for_irq_injection(vcpu, kvm_run)) {
-				++vcpu->stat.request_irq_exits;
-				post_kvm_run_save(vcpu, kvm_run);
+				r = -EINTR;
 				kvm_run->exit_reason = KVM_EXIT_INTR;
-				return -EINTR;
+				++vcpu->stat.request_irq_exits;
+				goto out;
+			}
+			if (!need_resched()) {
+				++vcpu->stat.light_exits;
+				goto again;
 			}
-
-			kvm_resched(vcpu);
-			goto again;
 		}
 	}
 
+out:
+	/*
+	 * Reload segment selectors ASAP. (it's needed for a functional
+	 * kernel: x86 relies on having __KERNEL_PDA in %fs and x86_64
+	 * relies on having 0 in %gs for the CPU PDA to work.)
+	 */
+	if (fs_gs_ldt_reload_needed) {
+		load_ldt(ldt_sel);
+		load_fs(fs_sel);
+		/*
+		 * If we have to reload gs, we must take care to
+		 * preserve our gs base.
+		 */
+		local_irq_disable();
+		load_gs(gs_sel);
+#ifdef CONFIG_X86_64
+		wrmsrl(MSR_GS_BASE, vmcs_readl(HOST_GS_BASE));
+#endif
+		local_irq_enable();
+
+		reload_tss();
+	}
+#ifdef CONFIG_X86_64
+	if (is_long_mode(vcpu)) {
+		save_msrs(vcpu->guest_msrs, NR_BAD_MSRS);
+		load_msrs(vcpu->host_msrs, NR_BAD_MSRS);
+	}
+#endif
+
+	if (r > 0) {
+		kvm_resched(vcpu);
+		goto preempted;
+	}
+
 	post_kvm_run_save(vcpu, kvm_run);
 	return r;
 }