On 2013-12-09 07:55, Gleb Natapov wrote:
> On Sun, Dec 08, 2013 at 02:00:11PM +0100, Jan Kiszka wrote:
>> On 2013-11-27 17:58, Paolo Bonzini wrote:
>>> Il 25/11/2013 14:37, Gleb Natapov ha scritto:
>>>> VM_(ENTRY|EXIT)_CONTROLS vmcs fields are read/written on each guest
>>>> entry but most times it can be avoided since values do not changes.
>>>> Keep fields copy in memory to avoid unnecessary reads from vmcs.
>>>>
>>>> Signed-off-by: Gleb Natapov <gleb@xxxxxxxxxx>
>>>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>>>> index b2fe1c2..29c1c7f 100644
>>>> --- a/arch/x86/kvm/vmx.c
>>>> +++ b/arch/x86/kvm/vmx.c
>>>> @@ -418,6 +418,8 @@ struct vcpu_vmx {
>>>> u64 msr_host_kernel_gs_base;
>>>> u64 msr_guest_kernel_gs_base;
>>>> #endif
>>>> + u32 vm_entry_controls_shadow;
>>>> + u32 vm_exit_controls_shadow;
>>>> /*
>>>> * loaded_vmcs points to the VMCS currently used in this vcpu. For a
>>>> * non-nested (L1) guest, it always points to vmcs01. For a nested
>>>> @@ -1326,6 +1328,62 @@ static void vmcs_set_bits(unsigned long field, u32 mask)
>>>> vmcs_writel(field, vmcs_readl(field) | mask);
>>>> }
>>>>
>>>> +static inline void vm_entry_controls_init(struct vcpu_vmx *vmx, u32 val)
>>>> +{
>>>> + vmcs_write32(VM_ENTRY_CONTROLS, val);
>>>> + vmx->vm_entry_controls_shadow = val;
>>>> +}
>>>> +
>>>> +static inline void vm_entry_controls_set(struct vcpu_vmx *vmx, u32 val)
>>>> +{
>>>> + if (vmx->vm_entry_controls_shadow != val)
>>>> + vm_entry_controls_init(vmx, val);
>>>> +}
>>>> +
>>>> +static inline u32 vm_entry_controls_get(struct vcpu_vmx *vmx)
>>>> +{
>>>> + return vmx->vm_entry_controls_shadow;
>>>> +}
>>>> +
>>>> +
>>>> +static inline void vm_entry_controls_setbit(struct vcpu_vmx *vmx, u32 val)
>>>> +{
>>>> + vm_entry_controls_set(vmx, vm_entry_controls_get(vmx) | val);
>>>> +}
>>>> +
>>>> +static inline void vm_entry_controls_clearbit(struct vcpu_vmx *vmx, u32 val)
>>>> +{
>>>> + vm_entry_controls_set(vmx, vm_entry_controls_get(vmx) & ~val);
>>>> +}
>>>> +
>>>> +static inline void vm_exit_controls_init(struct vcpu_vmx *vmx, u32 val)
>>>> +{
>>>> + vmcs_write32(VM_EXIT_CONTROLS, val);
>>>> + vmx->vm_exit_controls_shadow = val;
>>>> +}
>>>> +
>>>> +static inline void vm_exit_controls_set(struct vcpu_vmx *vmx, u32 val)
>>>> +{
>>>> + if (vmx->vm_exit_controls_shadow != val)
>>>> + vm_exit_controls_init(vmx, val);
>>>> +}
>>>> +
>>>> +static inline u32 vm_exit_controls_get(struct vcpu_vmx *vmx)
>>>> +{
>>>> + return vmx->vm_exit_controls_shadow;
>>>> +}
>>>> +
>>>> +
>>>> +static inline void vm_exit_controls_setbit(struct vcpu_vmx *vmx, u32 val)
>>>> +{
>>>> + vm_exit_controls_set(vmx, vm_exit_controls_get(vmx) | val);
>>>> +}
>>>> +
>>>> +static inline void vm_exit_controls_clearbit(struct vcpu_vmx *vmx, u32 val)
>>>> +{
>>>> + vm_exit_controls_set(vmx, vm_exit_controls_get(vmx) & ~val);
>>>> +}
>>>> +
>>>> static void vmx_segment_cache_clear(struct vcpu_vmx *vmx)
>>>> {
>>>> vmx->segment_cache.bitmask = 0;
>>>> @@ -1410,11 +1468,11 @@ static void update_exception_bitmap(struct kvm_vcpu *vcpu)
>>>> vmcs_write32(EXCEPTION_BITMAP, eb);
>>>> }
>>>>
>>>> -static void clear_atomic_switch_msr_special(unsigned long entry,
>>>> - unsigned long exit)
>>>> +static void clear_atomic_switch_msr_special(struct vcpu_vmx *vmx,
>>>> + unsigned long entry, unsigned long exit)
>>>> {
>>>> - vmcs_clear_bits(VM_ENTRY_CONTROLS, entry);
>>>> - vmcs_clear_bits(VM_EXIT_CONTROLS, exit);
>>>> + vm_entry_controls_clearbit(vmx, entry);
>>>> + vm_exit_controls_clearbit(vmx, exit);
>>>> }
>>>>
>>>> static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr)
>>>> @@ -1425,14 +1483,15 @@ static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr)
>>>> switch (msr) {
>>>> case MSR_EFER:
>>>> if (cpu_has_load_ia32_efer) {
>>>> - clear_atomic_switch_msr_special(VM_ENTRY_LOAD_IA32_EFER,
>>>> + clear_atomic_switch_msr_special(vmx,
>>>> + VM_ENTRY_LOAD_IA32_EFER,
>>>> VM_EXIT_LOAD_IA32_EFER);
>>>> return;
>>>> }
>>>> break;
>>>> case MSR_CORE_PERF_GLOBAL_CTRL:
>>>> if (cpu_has_load_perf_global_ctrl) {
>>>> - clear_atomic_switch_msr_special(
>>>> + clear_atomic_switch_msr_special(vmx,
>>>> VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL,
>>>> VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL);
>>>> return;
>>>> @@ -1453,14 +1512,15 @@ static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr)
>>>> vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->nr);
>>>> }
>>>>
>>>> -static void add_atomic_switch_msr_special(unsigned long entry,
>>>> - unsigned long exit, unsigned long guest_val_vmcs,
>>>> - unsigned long host_val_vmcs, u64 guest_val, u64 host_val)
>>>> +static void add_atomic_switch_msr_special(struct vcpu_vmx *vmx,
>>>> + unsigned long entry, unsigned long exit,
>>>> + unsigned long guest_val_vmcs, unsigned long host_val_vmcs,
>>>> + u64 guest_val, u64 host_val)
>>>> {
>>>> vmcs_write64(guest_val_vmcs, guest_val);
>>>> vmcs_write64(host_val_vmcs, host_val);
>>>> - vmcs_set_bits(VM_ENTRY_CONTROLS, entry);
>>>> - vmcs_set_bits(VM_EXIT_CONTROLS, exit);
>>>> + vm_entry_controls_setbit(vmx, entry);
>>>> + vm_exit_controls_setbit(vmx, exit);
>>>> }
>>>>
>>>> static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr,
>>>> @@ -1472,7 +1532,8 @@ static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr,
>>>> switch (msr) {
>>>> case MSR_EFER:
>>>> if (cpu_has_load_ia32_efer) {
>>>> - add_atomic_switch_msr_special(VM_ENTRY_LOAD_IA32_EFER,
>>>> + add_atomic_switch_msr_special(vmx,
>>>> + VM_ENTRY_LOAD_IA32_EFER,
>>>> VM_EXIT_LOAD_IA32_EFER,
>>>> GUEST_IA32_EFER,
>>>> HOST_IA32_EFER,
>>>> @@ -1482,7 +1543,7 @@ static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr,
>>>> break;
>>>> case MSR_CORE_PERF_GLOBAL_CTRL:
>>>> if (cpu_has_load_perf_global_ctrl) {
>>>> - add_atomic_switch_msr_special(
>>>> + add_atomic_switch_msr_special(vmx,
>>>> VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL,
>>>> VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL,
>>>> GUEST_IA32_PERF_GLOBAL_CTRL,
>>>> @@ -3182,14 +3243,10 @@ static void vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer)
>>>> vmx_load_host_state(to_vmx(vcpu));
>>>> vcpu->arch.efer = efer;
>>>> if (efer & EFER_LMA) {
>>>> - vmcs_write32(VM_ENTRY_CONTROLS,
>>>> - vmcs_read32(VM_ENTRY_CONTROLS) |
>>>> - VM_ENTRY_IA32E_MODE);
>>>> + vm_entry_controls_setbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE);
>>>> msr->data = efer;
>>>> } else {
>>>> - vmcs_write32(VM_ENTRY_CONTROLS,
>>>> - vmcs_read32(VM_ENTRY_CONTROLS) &
>>>> - ~VM_ENTRY_IA32E_MODE);
>>>> + vm_entry_controls_clearbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE);
>>>>
>>>> msr->data = efer & ~EFER_LME;
>>>> }
>>>> @@ -3217,9 +3274,7 @@ static void enter_lmode(struct kvm_vcpu *vcpu)
>>>>
>>>> static void exit_lmode(struct kvm_vcpu *vcpu)
>>>> {
>>>> - vmcs_write32(VM_ENTRY_CONTROLS,
>>>> - vmcs_read32(VM_ENTRY_CONTROLS)
>>>> - & ~VM_ENTRY_IA32E_MODE);
>>>> + vm_entry_controls_clearbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE);
>>>> vmx_set_efer(vcpu, vcpu->arch.efer & ~EFER_LMA);
>>>> }
>>>>
>>>> @@ -4346,10 +4401,11 @@ static int vmx_vcpu_setup(struct vcpu_vmx *vmx)
>>>> ++vmx->nmsrs;
>>>> }
>>>>
>>>> - vmcs_write32(VM_EXIT_CONTROLS, vmcs_config.vmexit_ctrl);
>>>> +
>>>> + vm_exit_controls_init(vmx, vmcs_config.vmexit_ctrl);
>>>>
>>>> /* 22.2.1, 20.8.1 */
>>>> - vmcs_write32(VM_ENTRY_CONTROLS, vmcs_config.vmentry_ctrl);
>>>> + vm_entry_controls_init(vmx, vmcs_config.vmentry_ctrl);
>>>>
>>>> vmcs_writel(CR0_GUEST_HOST_MASK, ~0UL);
>>>> set_cr4_guest_host_mask(vmx);
>>>> @@ -7759,12 +7815,12 @@ static void prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
>>>> exit_control = vmcs_config.vmexit_ctrl;
>>>> if (vmcs12->pin_based_vm_exec_control & PIN_BASED_VMX_PREEMPTION_TIMER)
>>>> exit_control |= VM_EXIT_SAVE_VMX_PREEMPTION_TIMER;
>>>> - vmcs_write32(VM_EXIT_CONTROLS, exit_control);
>>>> + vm_exit_controls_init(vmx, exit_control);
>>>>
>>>> /* vmcs12's VM_ENTRY_LOAD_IA32_EFER and VM_ENTRY_IA32E_MODE are
>>>> * emulated by vmx_set_efer(), below.
>>>> */
>>>> - vmcs_write32(VM_ENTRY_CONTROLS,
>>>> + vm_entry_controls_init(vmx,
>>>> (vmcs12->vm_entry_controls & ~VM_ENTRY_LOAD_IA32_EFER &
>>>> ~VM_ENTRY_IA32E_MODE) |
>>>> (vmcs_config.vmentry_ctrl & ~VM_ENTRY_IA32E_MODE));
>>>> @@ -8186,7 +8242,7 @@ static void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
>>>>
>>>> vmcs12->vm_entry_controls =
>>>> (vmcs12->vm_entry_controls & ~VM_ENTRY_IA32E_MODE) |
>>>> - (vmcs_read32(VM_ENTRY_CONTROLS) & VM_ENTRY_IA32E_MODE);
>>>> + (vm_entry_controls_get(to_vmx(vcpu)) & VM_ENTRY_IA32E_MODE);
>>>>
>>>> /* TODO: These cannot have changed unless we have MSR bitmaps and
>>>> * the relevant bit asks not to trap the change */
>>>> @@ -8390,6 +8446,7 @@ static void nested_vmx_vmexit(struct kvm_vcpu *vcpu)
>>>> vcpu->cpu = cpu;
>>>> put_cpu();
>>>>
>>>> + vm_entry_controls_init(vmx, vmcs_read32(VM_ENTRY_CONTROLS));
>>>> vmx_segment_cache_clear(vmx);
>>>>
>>>> /* if no vmcs02 cache requested, remove the one we used */
>>>> --
>>>> Gleb.
>>>> --
>>>> To unsubscribe from this list: send the line "unsubscribe kvm" in
>>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>>>
>>>
>>> Applied to kvm/queue, thanks.
>>
>> Seems to have or cause an issue with nVMX. The (still not merged) vmx
>> unit test breaks here after apply this commit:
>>
> What about this patch?
>
>
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 1d95384..f90320b 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -8452,6 +8452,7 @@ static void nested_vmx_vmexit(struct kvm_vcpu *vcpu)
> put_cpu();
>
> vm_entry_controls_init(vmx, vmcs_read32(VM_ENTRY_CONTROLS));
> + vm_exit_controls_init(vmx, vmcs_read32(VM_EXIT_CONTROLS));
> vmx_segment_cache_clear(vmx);
>
> /* if no vmcs02 cache requested, remove the one we used */
> --
> Gleb.
>
Yep, works and makes sense as well.
Thanks,
Jan
Attachment:
signature.asc
Description: OpenPGP digital signature
