public inbox for kvm@vger.kernel.org
 help / color / mirror / Atom feed
From: Avi Kivity <avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
To: kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
Cc: linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
	Qing He <qing.he-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Subject: [PATCH 079/104] KVM: deliver PIC interrupt only to vcpu0
Date: Mon, 17 Sep 2007 10:32:01 +0200	[thread overview]
Message-ID: <11900179493098-git-send-email-avi@qumranet.com> (raw)
In-Reply-To: <11900179463203-git-send-email-avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>

From: Qing He <qing.he-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

This patch changes the PIC interrupts delivery. Now it is only delivered
to vcpu0 when either condition is met (on vcpu0):
  1. local APIC is hardware disabled
  2. LVT0 is unmasked and configured to delivery mode ExtInt

It fixes the 2x faster wall clock on x86_64 and SMP i386 Linux guests

Signed-off-by: Eddie (Yaozu) Dong <eddie.dong-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Qing He <qing.he-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Avi Kivity <avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
---
 drivers/kvm/irq.c   |   15 ++++++++++-----
 drivers/kvm/irq.h   |    1 +
 drivers/kvm/lapic.c |   17 +++++++++++++++++
 3 files changed, 28 insertions(+), 5 deletions(-)

diff --git a/drivers/kvm/irq.c b/drivers/kvm/irq.c
index b88e501..7628c7f 100644
--- a/drivers/kvm/irq.c
+++ b/drivers/kvm/irq.c
@@ -33,8 +33,11 @@ int kvm_cpu_has_interrupt(struct kvm_vcpu *v)
 	struct kvm_pic *s;
 
 	if (kvm_apic_has_interrupt(v) == -1) {	/* LAPIC */
-		s = pic_irqchip(v->kvm);	/* PIC */
-		return s->output;
+		if (kvm_apic_accept_pic_intr(v)) {
+			s = pic_irqchip(v->kvm);	/* PIC */
+			return s->output;
+		} else
+			return 0;
 	}
 	return 1;
 }
@@ -50,9 +53,11 @@ int kvm_cpu_get_interrupt(struct kvm_vcpu *v)
 
 	vector = kvm_get_apic_interrupt(v);	/* APIC */
 	if (vector == -1) {
-		s = pic_irqchip(v->kvm);
-		s->output = 0;		/* PIC */
-		vector = kvm_pic_read_irq(s);
+		if (kvm_apic_accept_pic_intr(v)) {
+			s = pic_irqchip(v->kvm);
+			s->output = 0;		/* PIC */
+			vector = kvm_pic_read_irq(s);
+		}
 	}
 	return vector;
 }
diff --git a/drivers/kvm/irq.h b/drivers/kvm/irq.h
index f324cfb..ec46a09 100644
--- a/drivers/kvm/irq.h
+++ b/drivers/kvm/irq.h
@@ -135,6 +135,7 @@ do {									\
 
 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 int kvm_apic_has_interrupt(struct kvm_vcpu *vcpu);
+int kvm_apic_accept_pic_intr(struct kvm_vcpu *vcpu);
 int kvm_get_apic_interrupt(struct kvm_vcpu *vcpu);
 int kvm_create_lapic(struct kvm_vcpu *vcpu);
 void kvm_free_apic(struct kvm_lapic *apic);
diff --git a/drivers/kvm/lapic.c b/drivers/kvm/lapic.c
index 2706ec3..01e7696 100644
--- a/drivers/kvm/lapic.c
+++ b/drivers/kvm/lapic.c
@@ -788,6 +788,8 @@ static void lapic_reset(struct kvm_vcpu *vcpu)
 
 	for (i = 0; i < APIC_LVT_NUM; i++)
 		apic_set_reg(apic, APIC_LVTT + 0x10 * i, APIC_LVT_MASKED);
+	apic_set_reg(apic, APIC_LVT0,
+		     SET_APIC_DELIVERY_MODE(0, APIC_MODE_EXTINT));
 
 	apic_set_reg(apic, APIC_DFR, 0xffffffffU);
 	apic_set_reg(apic, APIC_SPIV, 0xff);
@@ -932,6 +934,21 @@ int kvm_apic_has_interrupt(struct kvm_vcpu *vcpu)
 	return highest_irr;
 }
 
+int kvm_apic_accept_pic_intr(struct kvm_vcpu *vcpu)
+{
+	u32 lvt0 = apic_get_reg(vcpu->apic, APIC_LVT0);
+	int r = 0;
+
+	if (vcpu->vcpu_id == 0) {
+		if (!apic_hw_enabled(vcpu->apic))
+			r = 1;
+		if ((lvt0 & APIC_LVT_MASKED) == 0 &&
+		    GET_APIC_DELIVERY_MODE(lvt0) == APIC_MODE_EXTINT)
+			r = 1;
+	}
+	return r;
+}
+
 void kvm_inject_apic_timer_irqs(struct kvm_vcpu *vcpu)
 {
 	struct kvm_lapic *apic = vcpu->apic;
-- 
1.5.3


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

  parent reply	other threads:[~2007-09-17  8:32 UTC|newest]

Thread overview: 114+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-09-17  8:30 [PATCH 000/104] KVM patch queue for the 2.6.24 merge window Avi Kivity
     [not found] ` <11900179463203-git-send-email-avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2007-09-17  8:30   ` [PATCH 001/104] KVM: Fix *nopage() in kvm_main.c Avi Kivity
     [not found]     ` <11900179461472-git-send-email-avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2007-09-17  9:13       ` Christoph Hellwig
     [not found]         ` <20070917091340.GB2083-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
2007-09-17  9:15           ` Avi Kivity
     [not found]             ` <46EE45C3.8040302-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2007-09-17  9:18               ` Avi Kivity
     [not found]                 ` <46EE4683.7070205-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2007-09-16 21:29                   ` Nick Piggin
     [not found]                     ` <200709170729.24835.nickpiggin-/E1597aS9LT0CCvOHzKKcA@public.gmane.org>
2007-09-17 18:19                       ` Avi Kivity
     [not found]                         ` <46EEC52B.8080101-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2007-09-17 17:17                           ` Nick Piggin
     [not found]                             ` <200709180317.07055.nickpiggin-/E1597aS9LT0CCvOHzKKcA@public.gmane.org>
2007-09-18 10:44                               ` Avi Kivity
2007-09-17  8:30   ` [PATCH 002/104] KVM: SMP: Add vcpu_id field in struct vcpu Avi Kivity
2007-09-17  8:30   ` [PATCH 003/104] KVM: Future-proof the exit information union ABI Avi Kivity
2007-09-17  8:30   ` [PATCH 004/104] KVM: In-kernel string pio write support Avi Kivity
2007-09-17  8:30   ` [PATCH 005/104] KVM: Trivial: /dev/kvm interface is no longer experimental Avi Kivity
2007-09-17  8:30   ` [PATCH 006/104] KVM: Trivial: Remove unused struct cpu_user_regs declaration Avi Kivity
2007-09-17  8:30   ` [PATCH 007/104] KVM: Trivial: Make decode_register() static Avi Kivity
2007-09-17  8:30   ` [PATCH 008/104] KVM: Trivial: Comment spelling may escape grep Avi Kivity
2007-09-17  8:30   ` [PATCH 009/104] KVM: Trivial: Avoid hardware_disable predeclaration Avi Kivity
2007-09-17  8:30   ` [PATCH 010/104] KVM: Trivial: Use standard CR0 flags macros from asm/cpu-features.h Avi Kivity
2007-09-17  8:30   ` [PATCH 011/104] KVM: Use standard CR3 flags, tighten checking Avi Kivity
2007-09-17  8:30   ` [PATCH 012/104] KVM: Use standard CR4 " Avi Kivity
2007-09-17  8:30   ` [PATCH 013/104] KVM: Trivial: Use standard BITMAP macros, open-code userspace-exposed header Avi Kivity
2007-09-17  8:30   ` [PATCH 014/104] KVM: Set exit_reason to KVM_EXIT_MMIO where run->mmio is initialized Avi Kivity
2007-09-17  8:30   ` [PATCH 015/104] KVM: Use standard CR8 flags, and fix TPR definition Avi Kivity
2007-09-17  8:30   ` [PATCH 016/104] KVM: x86 emulator: fix cmov for writeback changes Avi Kivity
2007-09-17  8:30   ` [PATCH 017/104] KVM: x86 emulator: fix faulty check for two-byte opcode Avi Kivity
2007-09-17  8:31   ` [PATCH 018/104] KVM: Return if the pdptrs are invalid when the guest turns on PAE Avi Kivity
2007-09-17  8:31   ` [PATCH 019/104] KVM: Hoist kvm_mmu_reload() out of the critical section Avi Kivity
2007-09-17  8:31   ` [PATCH 020/104] KVM: Move gfn_to_page out of kmap/unmap pairs Avi Kivity
2007-09-17  8:31   ` [PATCH 021/104] KVM: VMX: Import some constants of vmcs from IA32 SDM Avi Kivity
2007-09-17  8:31   ` [PATCH 022/104] KVM: Remove dead code in the cmpxchg instruction emulation Avi Kivity
2007-09-17  8:31   ` [PATCH 023/104] KVM: load_pdptrs() cleanups Avi Kivity
2007-09-17  8:31   ` [PATCH 024/104] KVM: Remove arch specific components from the general code Avi Kivity
2007-09-17  8:31   ` [PATCH 025/104] KVM: Dynamically allocate vcpus Avi Kivity
2007-09-17  8:31   ` [PATCH 026/104] KVM: VMX: Improve the method of writing vmcs control Avi Kivity
2007-09-17  8:31   ` [PATCH 027/104] KVM: add hypercall nr to kvm_run Avi Kivity
2007-09-17  8:31   ` [PATCH 028/104] KVM: Use the scheduler preemption notifiers to make kvm preemptible Avi Kivity
2007-09-17  8:31   ` [PATCH 029/104] KVM: Convert vm lock to a mutex Avi Kivity
2007-09-17  8:31   ` [PATCH 030/104] KVM: fx_init() needs preemption disabled while it plays with the FPU state Avi Kivity
2007-09-17  8:31   ` [PATCH 031/104] KVM: VMX: pass vcpu_vmx internally Avi Kivity
2007-09-17  8:31   ` [PATCH 032/104] KVM: Remove three magic numbers Avi Kivity
2007-09-17  8:31   ` [PATCH 033/104] KVM: SVM: de-containization Avi Kivity
2007-09-17  8:31   ` [PATCH 034/104] KVM: SVM: internal function name cleanup Avi Kivity
2007-09-17  8:31   ` [PATCH 035/104] KVM: Change the emulator_{read, write, cmpxchg}_* functions to take a vcpu Avi Kivity
2007-09-17  8:31   ` [PATCH 036/104] KVM: Remove kvm_{read,write}_guest() Avi Kivity
2007-09-17  8:31   ` [PATCH 037/104] KVM: Use kmem cache for allocating vcpus Avi Kivity
2007-09-17  8:31   ` [PATCH 038/104] KVM: Use alignment properties of vcpu to simplify FPU ops Avi Kivity
2007-09-17  8:31   ` [PATCH 039/104] KVM: kvm_vm_ioctl_get_dirty_log restore "nothing dirty" optimization Avi Kivity
2007-09-17  8:31   ` [PATCH 040/104] KVM: VMX: Add cpu consistency check Avi Kivity
2007-09-17  8:31   ` [PATCH 041/104] KVM: Don't assign vcpu->cr3 if it's invalid: check first, set last Avi Kivity
2007-09-17  8:31   ` [PATCH 042/104] KVM: Cleanup mark_page_dirty Avi Kivity
2007-09-17  8:31   ` [PATCH 043/104] KVM: SVM: Make set_msr_interception more reliable Avi Kivity
2007-09-17  8:31   ` [PATCH 044/104] KVM: Remove redundant alloc_vmcs_cpu declaration Avi Kivity
2007-09-17  8:31   ` [PATCH 045/104] KVM: Fix defined but not used warning in drivers/kvm/vmx.c Avi Kivity
     [not found]     ` <11900179473956-git-send-email-avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2007-09-17 16:27       ` David Miller
     [not found]         ` <20070917.092708.112622378.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org>
2007-09-17 17:35           ` Avi Kivity
2007-09-17  8:31   ` [PATCH 046/104] KVM: Remove stat_set from debugfs Avi Kivity
2007-09-17  8:31   ` [PATCH 047/104] KVM: Remove unneeded kvm_dev_open and kvm_dev_release functions Avi Kivity
2007-09-17  8:31   ` [PATCH 048/104] KVM: Add and use pr_unimpl for standard formatting of unimplemented features Avi Kivity
     [not found]     ` <11900179481707-git-send-email-avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2007-09-17 16:16       ` Joe Perches
2007-09-17 23:08         ` Rusty Russell
2007-09-17  8:31   ` [PATCH 049/104] KVM: Use kmem_cache_free for kmem_cache_zalloc'ed objects Avi Kivity
2007-09-17  8:31   ` [PATCH 050/104] KVM: VMX: Remove a duplicated ia32e mode vm entry control Avi Kivity
2007-09-17  8:31   ` [PATCH 051/104] KVM: Remove useless assignment Avi Kivity
2007-09-17  8:31   ` [PATCH 052/104] KVM: Cleanup string I/O instruction emulation Avi Kivity
2007-09-17  8:31   ` [PATCH 053/104] KVM: Clean up kvm_setup_pio() Avi Kivity
2007-09-17  8:31   ` [PATCH 055/104] KVM: Communicate cr8 changes to userspace Avi Kivity
2007-09-17  8:31   ` [PATCH 056/104] KVM: x86 emulator: implement 'and $imm, %{al|ax|eax}' Avi Kivity
2007-09-17  8:31   ` [PATCH 057/104] KVM: x86 emulator: implement 'jmp rel' instruction (opcode 0xe9) Avi Kivity
2007-09-17  8:31   ` [PATCH 058/104] KVM: x86 emulator: Implement 'jmp rel short' instruction (opcode 0xeb) Avi Kivity
2007-09-17  8:31   ` [PATCH 059/104] KVM: x86 emulator: implement 'push reg' (opcodes 0x50-0x57) Avi Kivity
2007-09-17  8:31   ` [PATCH 060/104] KVM: VMX: allow rmode_tss_base() to work with >2G of guest memory Avi Kivity
2007-09-17  8:31   ` [PATCH 061/104] KVM: Support more memory slots Avi Kivity
2007-09-17  8:31   ` [PATCH 062/104] KVM: X86 emulator: fix 'push reg' writeback Avi Kivity
2007-09-17  8:31   ` [PATCH 063/104] KVM: VMX: Split segments reload in vmx_load_host_state() Avi Kivity
2007-09-17  8:31   ` [PATCH 064/104] KVM: Add support for in-kernel PIC emulation Avi Kivity
2007-09-17  8:31   ` [PATCH 065/104] KVM: Define and use cr8 access functions Avi Kivity
2007-09-17  8:31   ` [PATCH 066/104] KVM: Emulate local APIC in kernel Avi Kivity
2007-09-17  8:31   ` [PATCH 067/104] KVM: In-kernel I/O APIC model Avi Kivity
2007-09-17  8:31   ` [PATCH 068/104] KVM: Emulate hlt in the kernel Avi Kivity
2007-09-17  8:31   ` [PATCH 069/104] KVM: Protect in-kernel pio using kvm->lock Avi Kivity
2007-09-17  8:31   ` [PATCH 070/104] KVM: Add get/set irqchip ioctls for in-kernel PIC live migration support Avi Kivity
2007-09-17  8:31   ` [PATCH 071/104] KVM: Bypass irq_pending get/set when using in kernel irqchip Avi Kivity
2007-09-17  8:31   ` [PATCH 072/104] KVM: in-kernel IOAPIC save and restore support Avi Kivity
2007-09-17  8:31   ` [PATCH 073/104] KVM: in-kernel LAPIC " Avi Kivity
2007-09-17  8:31   ` [PATCH 074/104] KVM: pending irq save/restore Avi Kivity
2007-09-17  8:31   ` [PATCH 075/104] KVM: VMX: Use shadow TPR/cr8 for 64-bits guests Avi Kivity
2007-09-17  8:31   ` [PATCH 076/104] KVM: Keep track of missed timer irq injections Avi Kivity
2007-09-17  8:31   ` [PATCH 077/104] KVM: Migrate lapic hrtimer when vcpu moves to another cpu Avi Kivity
2007-09-17  8:32   ` [PATCH 078/104] KVM: disable tpr/cr8 sync when in-kernel APIC is used Avi Kivity
2007-09-17  8:32   ` Avi Kivity [this message]
2007-09-17  8:32   ` [PATCH 080/104] KVM: round robin for APIC lowest priority delivery mode Avi Kivity
2007-09-17  8:32   ` [PATCH 081/104] KVM: enable in-kernel APIC INIT/SIPI handling Avi Kivity
2007-09-17  8:32   ` [PATCH 082/104] KVM: Set the ET flag in CR0 after initializing FX Avi Kivity
2007-09-17  8:32   ` [PATCH 083/104] KVM: Remove the unused invlpg member of struct kvm_arch_ops Avi Kivity
2007-09-17  8:32   ` [PATCH 084/104] KVM: Clean up unloved invlpg emulation Avi Kivity
2007-09-17  8:32   ` [PATCH 085/104] KVM: Keep control regs in sync Avi Kivity
2007-09-17  8:32   ` [PATCH 086/104] KVM: Hoist SVM's get_cs_db_l_bits into core code Avi Kivity
2007-09-17  8:32   ` [PATCH 088/104] KVM: Rename kvm_arch_ops to kvm_x86_ops Avi Kivity
2007-09-17  8:32   ` [PATCH 091/104] KVM: Move main vcpu loop into subarch independent code Avi Kivity
2007-09-17  8:32   ` [PATCH 092/104] KVM: VMX: Fix exit qualification width on i386 Avi Kivity
2007-09-17  8:32   ` [PATCH 093/104] KVM: x86 emulator: push imm8 Avi Kivity
2007-09-17  8:32   ` [PATCH 095/104] KVM: x86 emulator: pushf Avi Kivity
2007-09-17  8:32   ` [PATCH 096/104] KVM: Improve emulation failure reporting Avi Kivity
2007-09-17  8:32   ` [PATCH 097/104] KVM: x86 emulator: sort opcodes into ascending order Avi Kivity
2007-09-17  8:32   ` [PATCH 100/104] KVM: x86 emulator: lea Avi Kivity
2007-09-17  8:32   ` [PATCH 102/104] KVM: x86 emulator: fix src, dst value initialization Avi Kivity
2007-09-17  8:32   ` [PATCH 103/104] KVM: x86 emulator: popf Avi Kivity
2007-09-17  8:32 ` [PATCH 087/104] KVM: Simplify memory allocation Avi Kivity
2007-09-17  8:32 ` [PATCH 089/104] KVM: MMU: Don't do GFP_NOWAIT allocations Avi Kivity
2007-09-17  8:32 ` [PATCH 090/104] KVM: VMX: Move vm entry failure handling to the exit handler Avi Kivity
2007-09-17  8:32 ` [PATCH 094/104] KVM: x86 emulator: call near Avi Kivity
2007-09-17  8:32 ` [PATCH 098/104] KVM: x86 emulator: imlpement jump conditional relative Avi Kivity
2007-09-17  8:32 ` [PATCH 099/104] KVM: X86 emulator: jump conditional short Avi Kivity
2007-09-17  8:32 ` [PATCH 101/104] KVM: x86 emulator: jmp abs Avi Kivity

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=11900179493098-git-send-email-avi@qumranet.com \
    --to=avi-atkuwr5tajbwk0htik3j/w@public.gmane.org \
    --cc=kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org \
    --cc=linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
    --cc=qing.he-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox