From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gleb Natapov Subject: [PATCHv2] KVM: emulator: do not needlesly sync registers from emulator ctxt to vcpu Date: Thu, 31 Mar 2011 12:06:41 +0200 Message-ID: <20110331100641.GL7766@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: kvm@vger.kernel.org To: avi@redhat.com, mtosatti@redhat.com Return-path: Received: from mx1.redhat.com ([209.132.183.28]:43819 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751614Ab1CaKGn (ORCPT ); Thu, 31 Mar 2011 06:06:43 -0400 Received: from int-mx12.intmail.prod.int.phx2.redhat.com (int-mx12.intmail.prod.int.phx2.redhat.com [10.5.11.25]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id p2VA6g8H028604 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Thu, 31 Mar 2011 06:06:43 -0400 Content-Disposition: inline Sender: kvm-owner@vger.kernel.org List-ID: Currently we sync registers back and forth before/after exiting to userspace for IO, but during IO device model shouldn't need to read/write the registers, so we can as well skip those sync points. The only exaception is broken vmware backdor interface. The new code sync registers content during IO only if registers are read from/written to by userspace in the middle of the IO operation and this almost never happens in practise. Signed-off-by: Gleb Natapov --- v1->v2 - rebased onto see emulation patchset. diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h index e820c63..f68bacc 100644 --- a/arch/x86/include/asm/kvm_host.h +++ b/arch/x86/include/asm/kvm_host.h @@ -370,6 +370,8 @@ struct kvm_vcpu_arch { /* emulate context */ struct x86_emulate_ctxt emulate_ctxt; + bool emulate_regs_need_sync_to_vcpu; + bool emulate_regs_need_sync_from_vcpu; gpa_t time; struct pvclock_vcpu_time_info hv_clock; diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 9686547..eecad1b 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -4378,6 +4378,7 @@ static void init_emulate_ctxt(struct kvm_vcpu *vcpu) ? X86EMUL_MODE_PROT32 : X86EMUL_MODE_PROT16; memset(c, 0, sizeof(struct decode_cache)); memcpy(c->regs, vcpu->arch.regs, sizeof c->regs); + vcpu->arch.emulate_regs_need_sync_from_vcpu = false; } int kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq) @@ -4460,6 +4461,7 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, { int r; struct decode_cache *c = &vcpu->arch.emulate_ctxt.decode; + bool writeback = true; kvm_clear_exception_queue(vcpu); vcpu->arch.mmio_fault_cr2 = cr2; @@ -4500,9 +4502,12 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, return EMULATE_DONE; } - /* this is needed for vmware backdor interface to work since it + /* this is needed for vmware backdoor interface to work since it changes registers values during IO operation */ - memcpy(c->regs, vcpu->arch.regs, sizeof c->regs); + if (vcpu->arch.emulate_regs_need_sync_from_vcpu) { + vcpu->arch.emulate_regs_need_sync_from_vcpu = false; + memcpy(c->regs, vcpu->arch.regs, sizeof c->regs); + } restart: r = x86_emulate_insn(&vcpu->arch.emulate_ctxt); @@ -4520,19 +4525,28 @@ restart: } else if (vcpu->arch.pio.count) { if (!vcpu->arch.pio.in) vcpu->arch.pio.count = 0; + else + writeback = false; r = EMULATE_DO_MMIO; - } else if (vcpu->mmio_needed) + } else if (vcpu->mmio_needed) { + if (!vcpu->mmio_is_write) + writeback = false; r = EMULATE_DO_MMIO; - else if (r == EMULATION_RESTART) + } else if (r == EMULATION_RESTART) goto restart; else r = EMULATE_DONE; - toggle_interruptibility(vcpu, vcpu->arch.emulate_ctxt.interruptibility); - kvm_set_rflags(vcpu, vcpu->arch.emulate_ctxt.eflags); - kvm_make_request(KVM_REQ_EVENT, vcpu); - memcpy(vcpu->arch.regs, c->regs, sizeof c->regs); - kvm_rip_write(vcpu, vcpu->arch.emulate_ctxt.eip); + if (writeback) { + toggle_interruptibility(vcpu, + vcpu->arch.emulate_ctxt.interruptibility); + kvm_set_rflags(vcpu, vcpu->arch.emulate_ctxt.eflags); + kvm_make_request(KVM_REQ_EVENT, vcpu); + memcpy(vcpu->arch.regs, c->regs, sizeof c->regs); + vcpu->arch.emulate_regs_need_sync_to_vcpu = false; + kvm_rip_write(vcpu, vcpu->arch.emulate_ctxt.eip); + } else + vcpu->arch.emulate_regs_need_sync_to_vcpu = true; return r; } @@ -5516,6 +5530,18 @@ out: int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) { + if (vcpu->arch.emulate_regs_need_sync_to_vcpu) { + /* + * We are here if userspace calls get_regs() in the middle of + * instruction emulation. Registers state needs to be copied + * back from emulation context to vcpu. Usrapace shouldn't do + * that usually, but some bad designed PV devices (vmware + * backdoor interface) need this to work + */ + struct decode_cache *c = &vcpu->arch.emulate_ctxt.decode; + memcpy(vcpu->arch.regs, c->regs, sizeof c->regs); + vcpu->arch.emulate_regs_need_sync_to_vcpu = false; + } regs->rax = kvm_register_read(vcpu, VCPU_REGS_RAX); regs->rbx = kvm_register_read(vcpu, VCPU_REGS_RBX); regs->rcx = kvm_register_read(vcpu, VCPU_REGS_RCX); @@ -5543,6 +5569,9 @@ int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) { + vcpu->arch.emulate_regs_need_sync_from_vcpu = true; + vcpu->arch.emulate_regs_need_sync_to_vcpu = false; + kvm_register_write(vcpu, VCPU_REGS_RAX, regs->rax); kvm_register_write(vcpu, VCPU_REGS_RBX, regs->rbx); kvm_register_write(vcpu, VCPU_REGS_RCX, regs->rcx); -- Gleb.