From e9bda3b3d0ce775afe15eaf71922d342cc74991c Mon Sep 17 00:00:00 2001 From: Josh Triplett Date: Tue, 20 Mar 2012 23:33:51 -0700 Subject: KVM: VMX: Auto-load on CPUs with VMX Enable x86 feature-based autoloading for the kvm-intel module on CPUs with X86_FEATURE_VMX. Signed-off-by: Josh Triplett Acked-By: Kay Sievers Signed-off-by: Avi Kivity --- arch/x86/kvm/vmx.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'arch/x86/kvm/vmx.c') diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index ad85adfef84..52f68563576 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -27,6 +27,7 @@ #include #include #include +#include #include #include #include @@ -51,6 +52,12 @@ MODULE_AUTHOR("Qumranet"); MODULE_LICENSE("GPL"); +static const struct x86_cpu_id vmx_cpu_id[] = { + X86_FEATURE_MATCH(X86_FEATURE_VMX), + {} +}; +MODULE_DEVICE_TABLE(x86cpu, vmx_cpu_id); + static bool __read_mostly enable_vpid = 1; module_param_named(vpid, enable_vpid, bool, 0444); -- cgit v1.2.3-70-g09d2 From 5f3fbc342f408199e5cbb4b3dc220569147a99a7 Mon Sep 17 00:00:00 2001 From: Xiao Guangrong Date: Mon, 14 May 2012 14:58:58 +0800 Subject: KVM: VMX: unlike vmcs on fail path fix: [ 1529.577273] Call Trace: [ 1529.577289] [] kvm_arch_hardware_disable+0x13/0x30 [kvm] [ 1529.577302] [] hardware_disable_nolock+0x35/0x39 [kvm] [ 1529.577311] [] ? cpumask_clear_cpu.constprop.31+0x13/0x13 [kvm] [ 1529.577315] [] on_each_cpu+0x44/0x84 [ 1529.577326] [] hardware_disable_all_nolock+0x34/0x36 [kvm] [ 1529.577335] [] hardware_disable_all+0x2b/0x39 [kvm] [ 1529.577349] [] kvm_put_kvm+0xed/0x10f [kvm] [ 1529.577358] [] kvm_vm_release+0x22/0x28 [kvm] Signed-off-by: Xiao Guangrong Signed-off-by: Avi Kivity --- arch/x86/kvm/vmx.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kvm/vmx.c') diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index 61ebdb6390e..3062ea95266 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -6350,7 +6350,7 @@ static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id) return &vmx->vcpu; free_vmcs: - free_vmcs(vmx->loaded_vmcs->vmcs); + free_loaded_vmcs(vmx->loaded_vmcs); free_msrs: kfree(vmx->guest_msrs); uninit_vcpu: -- cgit v1.2.3-70-g09d2 From 512d5649e8dc3ed36f2ebf0818da64a4d4c2544a Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Sun, 13 May 2012 19:53:23 +0300 Subject: KVM: VMX: Fix %ds/%es clobber The vmx exit code unconditionally restores %ds and %es to __USER_DS. This can override the user's values, since %ds and %es are not saved and restored in x86_64 syscalls. In practice, this isn't dangerous since nobody uses segment registers in long mode, least of all programs that use KVM. Signed-off-by: Avi Kivity Signed-off-by: Marcelo Tosatti --- arch/x86/kvm/vmx.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'arch/x86/kvm/vmx.c') diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index 3062ea95266..f2ee016e100 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -6102,7 +6102,10 @@ static void atomic_switch_perf_msrs(struct vcpu_vmx *vmx) static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu) { struct vcpu_vmx *vmx = to_vmx(vcpu); + u16 _ds, _es; + savesegment(ds, _ds); + savesegment(es, _es); if (is_guest_mode(vcpu) && !vmx->nested.nested_run_pending) { struct vmcs12 *vmcs12 = get_vmcs12(vcpu); if (vmcs12->idt_vectoring_info_field & @@ -6263,7 +6266,8 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu) } } - asm("mov %0, %%ds; mov %0, %%es" : : "r"(__USER_DS)); + loadsegment(ds, _ds); + loadsegment(es, _es); vmx->loaded_vmcs->launched = 1; vmx->exit_reason = vmcs_read32(VM_EXIT_REASON); -- cgit v1.2.3-70-g09d2 From b2da15ac26a0c00fc0d399a2bc5cf3c4e15f0b4f Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Sun, 13 May 2012 19:53:24 +0300 Subject: KVM: VMX: Optimize %ds, %es reload On x86_64, we can defer %ds and %es reload to the heavyweight context switch, since nothing in the lightweight paths uses the host %ds or %es (they are ignored by the processor). Furthermore we can avoid the load if the segments are null, by letting the hardware load the null segments for us. This is the expected case. On i386, we could avoid the reload entirely, since the entry.S paths take care of reload, except for the SYSEXIT path which leaves %ds and %es set to __USER_DS. So we set them to the same values as well. Saves about 70 cycles out of 1600 (around 4%; noisy measurements). Signed-off-by: Avi Kivity Signed-off-by: Marcelo Tosatti --- arch/x86/kvm/vmx.c | 36 +++++++++++++++++++++++++++++++----- 1 file changed, 31 insertions(+), 5 deletions(-) (limited to 'arch/x86/kvm/vmx.c') diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index f2ee016e100..32eb5886629 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -393,6 +393,9 @@ struct vcpu_vmx { struct { int loaded; u16 fs_sel, gs_sel, ldt_sel; +#ifdef CONFIG_X86_64 + u16 ds_sel, es_sel; +#endif int gs_ldt_reload_needed; int fs_reload_needed; } host_state; @@ -1417,6 +1420,11 @@ static void vmx_save_host_state(struct kvm_vcpu *vcpu) vmx->host_state.gs_ldt_reload_needed = 1; } +#ifdef CONFIG_X86_64 + savesegment(ds, vmx->host_state.ds_sel); + savesegment(es, vmx->host_state.es_sel); +#endif + #ifdef CONFIG_X86_64 vmcs_writel(HOST_FS_BASE, read_msr(MSR_FS_BASE)); vmcs_writel(HOST_GS_BASE, read_msr(MSR_GS_BASE)); @@ -1457,6 +1465,19 @@ static void __vmx_load_host_state(struct vcpu_vmx *vmx) } if (vmx->host_state.fs_reload_needed) loadsegment(fs, vmx->host_state.fs_sel); +#ifdef CONFIG_X86_64 + if (unlikely(vmx->host_state.ds_sel | vmx->host_state.es_sel)) { + loadsegment(ds, vmx->host_state.ds_sel); + loadsegment(es, vmx->host_state.es_sel); + } +#else + /* + * The sysexit path does not restore ds/es, so we must set them to + * a reasonable value ourselves. + */ + loadsegment(ds, __USER_DS); + loadsegment(es, __USER_DS); +#endif reload_tss(); #ifdef CONFIG_X86_64 wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base); @@ -3640,8 +3661,18 @@ static void vmx_set_constant_host_state(void) vmcs_writel(HOST_CR3, read_cr3()); /* 22.2.3 FIXME: shadow tables */ vmcs_write16(HOST_CS_SELECTOR, __KERNEL_CS); /* 22.2.4 */ +#ifdef CONFIG_X86_64 + /* + * Load null selectors, so we can avoid reloading them in + * __vmx_load_host_state(), in case userspace uses the null selectors + * too (the expected case). + */ + vmcs_write16(HOST_DS_SELECTOR, 0); + vmcs_write16(HOST_ES_SELECTOR, 0); +#else vmcs_write16(HOST_DS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ vmcs_write16(HOST_ES_SELECTOR, __KERNEL_DS); /* 22.2.4 */ +#endif vmcs_write16(HOST_SS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ vmcs_write16(HOST_TR_SELECTOR, GDT_ENTRY_TSS*8); /* 22.2.4 */ @@ -6102,10 +6133,7 @@ static void atomic_switch_perf_msrs(struct vcpu_vmx *vmx) static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu) { struct vcpu_vmx *vmx = to_vmx(vcpu); - u16 _ds, _es; - savesegment(ds, _ds); - savesegment(es, _es); if (is_guest_mode(vcpu) && !vmx->nested.nested_run_pending) { struct vmcs12 *vmcs12 = get_vmcs12(vcpu); if (vmcs12->idt_vectoring_info_field & @@ -6266,8 +6294,6 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu) } } - loadsegment(ds, _ds); - loadsegment(es, _es); vmx->loaded_vmcs->launched = 1; vmx->exit_reason = vmcs_read32(VM_EXIT_REASON); -- cgit v1.2.3-70-g09d2