lkml.org 
[lkml]   [2021]   [Jan]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 1/2] KVM: x86: introduce definitions to support static calls for kvm_x86_ops
From
Date
On 1/12/21 6:04 PM, Sean Christopherson wrote:
> On Mon, Jan 11, 2021, Jason Baron wrote:
>> Use static calls to improve kvm_x86_ops performance. Introduce the
>> definitions that will be used by a subsequent patch to actualize the
>> savings.
>>
>> Note that all kvm_x86_ops are covered here except for 'pmu_ops' and
>> 'nested ops'. I think they can be covered by static calls in a simlilar
>> manner, but were omitted from this series to reduce scope and because
>> I don't think they have as large of a performance impact.
>>
>> Cc: Paolo Bonzini <pbonzini@redhat.com>
>> Cc: Thomas Gleixner <tglx@linutronix.de>
>> Cc: Ingo Molnar <mingo@redhat.com>
>> Cc: Borislav Petkov <bp@alien8.de>
>> Cc: Peter Zijlstra <peterz@infradead.org>
>> Cc: Andrea Arcangeli <aarcange@redhat.com>
>> Signed-off-by: Jason Baron <jbaron@akamai.com>
>> ---
>> arch/x86/include/asm/kvm_host.h | 65 +++++++++++++++++++++++++++++++++++++++++
>> arch/x86/kvm/x86.c | 5 ++++
>> 2 files changed, 70 insertions(+)
>>
>> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
>> index 3ab7b46..e947522 100644
>> --- a/arch/x86/include/asm/kvm_host.h
>> +++ b/arch/x86/include/asm/kvm_host.h
>> @@ -1087,6 +1087,65 @@ static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
>> return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
>> }
>>
>> +/*
>> + * static calls cover all kvm_x86_ops except for functions under pmu_ops and
>> + * nested_ops.
>> + */
>> +#define FOREACH_KVM_X86_OPS(F) \
>> + F(hardware_enable); F(hardware_disable); F(hardware_unsetup); \
>> + F(cpu_has_accelerated_tpr); F(has_emulated_msr); \
>> + F(vcpu_after_set_cpuid); F(vm_init); F(vm_destroy); F(vcpu_create); \
>> + F(vcpu_free); F(vcpu_reset); F(prepare_guest_switch); F(vcpu_load); \
>> + F(vcpu_put); F(update_exception_bitmap); F(get_msr); F(set_msr); \
>> + F(get_segment_base); F(get_segment); F(get_cpl); F(set_segment); \
>> + F(get_cs_db_l_bits); F(set_cr0); F(is_valid_cr4); F(set_cr4); \
>> + F(set_efer); F(get_idt); F(set_idt); F(get_gdt); F(set_gdt); \
>> + F(sync_dirty_debug_regs); F(set_dr7); F(cache_reg); F(get_rflags); \
>> + F(set_rflags); F(tlb_flush_all); F(tlb_flush_current); \
>> + F(tlb_remote_flush); F(tlb_remote_flush_with_range); F(tlb_flush_gva); \
>> + F(tlb_flush_guest); F(run); F(handle_exit); \
>> + F(skip_emulated_instruction); F(update_emulated_instruction); \
>> + F(set_interrupt_shadow); F(get_interrupt_shadow); F(patch_hypercall); \
>> + F(set_irq); F(set_nmi); F(queue_exception); F(cancel_injection); \
>> + F(interrupt_allowed); F(nmi_allowed); F(get_nmi_mask); F(set_nmi_mask);\
>> + F(enable_nmi_window); F(enable_irq_window); F(update_cr8_intercept); \
>> + F(check_apicv_inhibit_reasons); F(pre_update_apicv_exec_ctrl); \
>> + F(refresh_apicv_exec_ctrl); F(hwapic_irr_update); F(hwapic_isr_update);\
>> + F(guest_apic_has_interrupt); F(load_eoi_exitmap); \
>> + F(set_virtual_apic_mode); F(set_apic_access_page_addr); \
>> + F(deliver_posted_interrupt); F(sync_pir_to_irr); F(set_tss_addr); \
>> + F(set_identity_map_addr); F(get_mt_mask); F(load_mmu_pgd); \
>> + F(has_wbinvd_exit); F(write_l1_tsc_offset); F(get_exit_info); \
>> + F(check_intercept); F(handle_exit_irqoff); F(request_immediate_exit); \
>> + F(sched_in); F(slot_enable_log_dirty); F(slot_disable_log_dirty); \
>> + F(flush_log_dirty); F(enable_log_dirty_pt_masked); \
>> + F(cpu_dirty_log_size); F(pre_block); F(post_block); F(vcpu_blocking); \
>> + F(vcpu_unblocking); F(update_pi_irte); F(apicv_post_state_restore); \
>> + F(dy_apicv_has_pending_interrupt); F(set_hv_timer); F(cancel_hv_timer);\
>> + F(setup_mce); F(smi_allowed); F(pre_enter_smm); F(pre_leave_smm); \
>> + F(enable_smi_window); F(mem_enc_op); F(mem_enc_reg_region); \
>> + F(mem_enc_unreg_region); F(get_msr_feature); \
>> + F(can_emulate_instruction); F(apic_init_signal_blocked); \
>> + F(enable_direct_tlbflush); F(migrate_timers); F(msr_filter_changed); \
>> + F(complete_emulated_msr)
> What about adding a dedicated .h file for this beast? Then it won't be so
> painful to do one function per line. As is, updates to kvm_x86_ops will be
> messy.
Hi,

I'm fine moving it to a new header. I had put it right above
'struct kvm_x86_ops' in arch/x86/include/asm/kvm_host.h, such
that it would be updated along with kvm_x86_ops changes. But
I can put a big comment there to update the new header.

>
> And add yet another macro layer (or maybe just tweak this one?) so that the
> caller controls the line ending? I suppose you could also just use a comma, but
> that's a bit dirty...
>
> That would also allow using this to declare vmx_x86_ops and svm_x86_ops, which
> would need a comma insteat of a semi-colon. There have a been a few attempts to
> add a bit of automation to {vmx,svm}_x86_ops, this seems like it would be good
> motivation to go in a different direction and declare/define all ops, e.g. the
> VMX/SVM code could simply do something like:
>
> #define DECLARE_VMX_X86_OP(func) \
> .func = vmx_##func
>
> static struct kvm_x86_ops vmx_x86_ops __initdata = {
> .vm_size = sizeof(struct kvm_vmx),
> .vm_init = vmx_vm_init,
>
> .pmu_ops = &intel_pmu_ops,
> .nested_ops = &vmx_nested_ops,
>
> FOREACH_KVM_X86_OPS(DECLARE_VMX_X86_OP)
> };
>
Looking at the vmx definitions I see quite a few that don't
match that naming. For example:

hardware_unsetup,
hardware_enable,
hardware_disable,
report_flexpriority,
update_exception_bitmap,
enable_nmi_window,
enable_irq_window,
update_cr8_intercept,
pi_has_pending_interrupt,
cpu_has_vmx_wbinvd_exit,
pi_update_irte,
kvm_complete_insn_gp,

So I'm not sure if we want to extend these macros to
vmx/svm.

Thanks,

-Jason

\
 
 \ /
  Last update: 2021-01-13 05:14    [W:0.091 / U:0.172 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site