首先简单介绍下apic的destination mode：

physical mode下高32bit(destination field)是apic id

logical mode下高32bit是MDA，又分为flat和cluster两种：

flat就是一个bitmap

而cluster包含两个部分，高16bit为cluster ID，后16bit为bitmap；这里还分flat和hierarchy；这里就不深究了。

根据intel spec：Flat logicalmode is not supported in the x2APIC mode. Hence the Destination Format Register(DFR) is eliminated in x2APIC mode. 在x2apic已经没有FDR，所以在logical mode下不再支持flat mode了，只有cluster mode由于x2apic在原来apic的8bit apicid基础上扩展到32bit

所以通过ioapic发中断和发MSI(不是MSIX)中断的设备就有问题了；他们需要做IR，interrupt remapping(VT-d提供的一项能力，通过IOMMU来做)

但是，也有些例外，比如使用physical mode且cpu个数小于256时，8bit的apic id已经完全可以cover所有的cpu了，而且这8bit直接可以写入x2apic的32bit的destination field，所以这种情况下可以直接使用x2apic physical mode而无需IR能力

https://patchwork.kernel.org/patch/2826342/这个邮件内容告诉我们，使用x2apic而同时又没有IR能力的情况，只有虚拟化场景(bios不能enable VT-d这种情况好像要被禁止了，enable x2apic一定要开VT-d)

然而，虚拟化场景下为何一定要enable x2apic呢？

IBM给出了答案：Red Hat Enterprise Linux 6 implements x2APIC emulation for KVM guests. Advanced programmable interrupt controllers (APICs) are used in symmetric multiprocessor (SMP) computer systems. x2APIC is a machine state register (MSR) interface to a local APIC with performance and scalability enhancements. IBM lab tests show that enabling the x2APIC support for Red Hat Enterprise Linux 6 guests can result in 2% to 5% throughput improvement for many I/O workloads. You can enable the x2APIC support by specifying the -cpu qemu64,+x2apic option on the qemu-kvm command for a KVM guest.

然而这解释当然是隔靴搔痒，真正的解释还是要show me the patch：

diff

commit ce69a784504222c3ab6f1b3c357d09ec5772127a

Author: Gleb Natapov

Date: Mon Jul 20 15:24:17 2009 +0300

x86/apic: Enable x2APIC without interrupt remapping under KVM

KVM would like to provide x2APIC interface to a guest without emulating

interrupt remapping device. The reason KVM prefers guest to use x2APIC

is that x2APIC interface is better virtualizable and provides better

performance than mmio xAPIC interface:

- msr exits are faster than mmio (no page table walk, emulation)

- no need to read back ICR to look at the busy bit

- one 64 bit ICR write instead of two 32 bit writes

- shared code with the Hyper-V paravirt interface

Included patch changes x2APIC enabling logic to enable it even if IR

initialization failed, but kernel runs under KVM and no apic id is

greater than 255 (if there is one spec requires BIOS to move to x2apic

mode before starting an OS).

-v2: fix build

-v3: fix bug causing compiler warning

Signed-off-by: Gleb Natapov

Acked-by: Suresh Siddha

Cc: Sheng Yang

Cc: "avi@redhat.com"

LKML-Reference: <20090720122417.GR5638@redhat.com>

Signed-off-by: Ingo Molnar

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

commitce69a784504222c3ab6f1b3c357d09ec5772127a

Author:GlebNatapov

Date:MonJul2015:24:172009+0300

x86/apic: Enable x2APIC without interrupt remapping under KVM

KVM would like to provide x2APIC interface to a guest without emulating

interrupt remapping device. The reason KVM prefers guest to use x2APIC

is that x2APIC interface is better virtualizable and provides better

performance than mmio xAPIC interface:

- msr exits are faster than mmio (no page table walk, emulation)

- no need to read back ICR to look at the busy bit

- one 64 bit ICR write instead of two 32 bit writes

- shared code with the Hyper-V paravirt interface

Included patch changes x2APIC enabling logic to enable it even if IR

initialization failed, but kernel runs under KVM and no apic id is

greater than 255 (if there is one spec requires BIOS to move to x2apic

mode before starting an OS).

-v2: fix build

-v3: fix bug causing compiler warning

Signed-off-by: Gleb Natapov

Acked-by: Suresh Siddha

Cc: Sheng Yang

Cc: "avi@redhat.com"

LKML-Reference: <20090720122417.GR5638@redhat.com>

Signed-off-by: Ingo Molnar

Related