WO2016092667A1

WO2016092667A1 - Computer and interrupt control method

Info

Publication number: WO2016092667A1
Application number: PCT/JP2014/082833
Authority: WO
Inventors: 貴之今田; 直也服部; 俊臣森木
Original assignee: 株式会社日立製作所
Priority date: 2014-12-11
Filing date: 2014-12-11
Publication date: 2016-06-16

Abstract

In a computer in which a plurality of virtual machines (VMs) is operated, the plurality of VMs is managed in such a manner that the VMs are separated into a priority group constituted of VMs in a priority mode and a non-priority group constituted of VMs in a non-priority mode. A virtualization management unit sets a physical computation device except a physical computation device assigned to the VM in the priority mode as a physical computation device for receiving a hardware interrupt issued from a physical interrupt device assigned to a virtual interrupt device held by the VM in the priority mode. The virtualization management unit operated by the physical computation device that has received the hardware interrupt specifies a physical computation device assigned to the VM in the priority mode which is a transfer destination of the hardware interrupt, and transfers the hardware interrupt to the specified physical computation device. The physical computation device that has received the transferred hardware interrupt specifies the VM as a target of the hardware interrupt, and issues a virtual interrupt with respect to a virtual computation device held by the VM specified on the basis of the hardware interrupt.

Description

Computer and interrupt control method

The present invention relates to an interrupt control technique when virtual software is used in a computer having a multi-core CPU configuration.

In recent years, a single multi-core CPU has been divided into multiple logical partitions for the purpose of efficient use of physical resources (computer resources) such as CPU (Central Processing Unit), memory, and I / O devices, and simplification of operation management. Virtualization software that divides into two is generally used.

The main role of virtualization software is to manage the correspondence between physical resources and virtual machines. In order to manage physical resources, the virtualization software constantly monitors the usage status of CPU time on the virtual machine and the communication status due to interrupts, suppresses the use of physical resources exceeding the allocated amount, or substitutes for processing I do.

When the virtualization software executes all the processes as described above, a large performance drop (overhead) occurs. Therefore, in recent years, CPUs such as Intel Corporation have a dedicated support mechanism for supporting management of physical resources by virtualization software (for example, VT-x: Virtualization Technology for Xeon).

Virtualization software is mainly used for applications whose main factors are arithmetic processing and memory access. This is because in recent years, CPU vendors have mounted dedicated support mechanisms on CPUs, mainly for the purpose of speeding up arithmetic processing and memory access.

In the future, in order to realize further speedup of arithmetic processing and high-frequency I / O processing, it is necessary to shorten the processing time of the hardware interrupt processed by the virtualization software as much as possible.

JP 2010-128911 A

The current CPU distinguishes between an operation mode in which the guest OS on the virtual machine is executed (hereinafter referred to as non-root mode) and an operation mode in which the virtualization software itself is executed (hereinafter referred to as root mode). This is the mainstream method.

In a CPU adopting such a method, when a hardware interrupt occurs for a physical CPU core that is operating in a virtual CPU core in non-root mode, (1) transition of the physical CPU core operation mode to root mode, (2) Hardware arbitration processing by virtualization software, (3) Transition of physical CPU core operation mode to non-root mode and transmission of hardware interrupt to virtual machine, (4) Interrupt handler in guest OS A series of processing called processing is required. In order to reduce the interrupt processing time in the server virtualization environment, it is particularly important to reduce the processing time from (1) to (3).

Also, the recent hardware support function has a mechanism for directly transmitting a hardware interrupt to the virtual machine without changing the CPU operation mode in order to reduce (1) to (4).

However, in order to use such a mechanism, it is necessary to process a hardware interrupt to the CPU core on the guest OS. In such a configuration, an interrupt that should originally be received by the virtualization software may be masked by a malicious guest OS. In this case, there is a possibility that the virtualization software cannot operate normally.

In the present invention, in a computer having a multi-core CPU, virtualization software, and a plurality of virtual machines managed by the virtualization software, the hardware interrupt processing in the virtual machines is accelerated, and the security of the virtualization software is ensured. For the purpose.

A typical example of the invention disclosed in the present application is as follows. That is, a computer on which a plurality of virtual machines generated using logically divided computer resources operate, and the computer resources include a plurality of physical operation devices, one or more physical memories, and the physical operation Including one or more physical interrupt devices that generate hardware interrupts to a device, each of the plurality of virtual machines includes a plurality of virtual arithmetic devices, virtual memories, and virtual interrupt devices, One or more physical memories generate a plurality of virtual machines, store a program that realizes a virtualization management unit that manages the plurality of virtual machines, and control the virtual machines on the plurality of virtual machines Operating the operating system, and the plurality of physical computing devices include a first operation mode for executing processing by the virtualization management unit and the operating system. A virtualization support unit that switches between the second operation mode for executing the processing and interrupt virtualization for transferring the hardware interrupt issued from the one or more physical interrupt devices between the physical arithmetic devices A priority group composed of virtual machines in a priority mode that is a control mode for avoiding a decrease in processing performance of the operating system due to the hardware interrupt, Managing the plurality of virtual machines in a non-priority group composed of virtual machines in non-priority mode, which is a control mode that allows a reduction in processing performance of the operating system due to the hardware interrupt, Before the virtual computer in the priority mode has a physical arithmetic device other than the physical arithmetic device assigned to the virtual computer in the priority mode Set as a physical arithmetic device that receives the hardware interrupt issued from the one or more physical interrupt devices assigned to the virtual interrupt device, and receive the hardware interrupt for the virtual machine in the priority mode The virtualization management unit executed by the physical arithmetic device specifies the physical arithmetic device assigned to the virtual machine in the priority mode as a transfer destination of the received hardware interrupt, and specifies the specified physical operation The physical processing device that transferred the received hardware interrupt to the device and received the transferred hardware interrupt specifies the virtual machine that is the target of the transferred hardware interrupt, and the transferred Issuing a virtual interrupt to the virtual arithmetic unit of the identified virtual machine based on the hardware interrupt And

According to the present invention, the hardware interrupt for the priority mode virtual machine is received by an arithmetic unit other than the arithmetic unit assigned to the priority mode virtual machine. Therefore, it is possible to avoid the overhead of the operating system processing on the virtual machine in the priority mode, and to speed up the hardware interrupt processing. Furthermore, since all hardware interrupts are processed by the virtualization manager, the security of the virtualization software can be ensured.

Issues, configurations, and effects other than those described above will be clarified by the description of the following examples.

FIG. 3 is a block diagram illustrating an example of a configuration of a physical computer according to the first embodiment. FIG. 6 is an explanatory diagram illustrating an example of a physical vector number management table held by a virtualization management unit according to the first embodiment. FIG. 6 is an explanatory diagram illustrating an example of a virtual vector number management table held by a VM according to the first embodiment. It is explanatory drawing which shows an example of the CPU interruption identifier management table which the virtualization management part of Example 1 hold | maintains. 6 is an explanatory diagram illustrating an example of a VM interrupt device management table held by a virtualization management unit according to the first embodiment. FIG. 6 is an explanatory diagram illustrating an example of a VM attribute table held by a virtualization management unit according to the first embodiment. FIG. FIG. 6 is an explanatory diagram illustrating an example of a physical CPU core group management table held by a virtualization management unit according to the first embodiment. It is explanatory drawing which shows an example of the high-speed IPI control data which the virtualization management part of Example 1 hold | maintains. 10 is a flowchart illustrating an example of a VM attribute table setting process executed by a virtualization management unit according to the first embodiment. 6 is a flowchart illustrating an example of a physical CPU core group management table update process executed by a virtualization management unit according to the first embodiment. 6 is a flowchart illustrating an example of an interrupt setting process for a guest OS executed by a virtualization management unit according to the first embodiment. 6 is a flowchart illustrating processing executed by the virtualization management unit according to the first embodiment during virtual CPU core scheduling. 6 is a flowchart illustrating an example of a startup process executed by a virtualization management unit according to the first embodiment. 6 is a flowchart illustrating an example of a startup process executed by a virtualization management unit according to the first embodiment. 10 is a flowchart illustrating an example of an interrupt setting process for a guest OS executed by a virtualization management unit according to the second embodiment. 12 is a flowchart illustrating an example of a startup process executed by a virtualization management unit according to the second embodiment.

Hereinafter, embodiments will be described with reference to the accompanying drawings.

In the first embodiment, processing attributes are set for each of two virtual machines generated on one physical computer. Specifically, “guest OS processing priority mode” is set for one virtual machine, and “guest OS processing non-priority mode” is set for one virtual machine.

In the case of hardware interrupt for a virtual machine in “guest OS processing priority mode”, the hardware issued from the interrupt device to the physical CPU core assigned to the virtual CPU core of the virtual machine in “guest OS processing non-priority mode” Receive an interrupt. Further, the physical CPU core that has received the hardware interrupt transfers the hardware interrupt to the physical CPU core assigned to the virtual CPU core of the virtual machine in the “guest OS processing priority mode”.

The physical CPU core assigned to the virtual CPU core of the virtual machine in the “guest OS processing priority mode” does not transition to the root mode. Therefore, a decrease in processing performance is suppressed in the virtual machine in “guest OS processing priority mode”. On the other hand, in the case of hardware interrupt for a virtual machine in the “guest OS processing non-priority mode”, the physical CPU core assigned to the virtual CPU core of the virtual machine receives the hardware interrupt from the interrupt device.

In the following explanation, hardware interrupts are also referred to as HW interrupts.

FIG. 1 is a block diagram illustrating an example of the configuration of the physical computer 10 according to the first embodiment. Hereinafter, each of the hardware configuration and the software configuration of the physical computer 10 will be described.

(Hardware configuration)
First, the hardware configuration of the physical computer 10 will be described with reference to FIG.

The physical computer 10 includes a CPU resource 100, a memory resource 110, and an interrupt device resource 115 as computer resources.

The CPU resource 100 is composed of a plurality of physical CPU cores 101 using a multiprocessor configuration or a multicore CPU. Each physical CPU core 101 includes a virtualization support function unit (not shown) (for example, Intel VT-x) and an interrupt virtualization support (Posted Interrupt Processing) used when the virtualization support function unit is enabled. A function unit 102 is included. The interrupt virtualization support function unit 102 is described in Non-Patent Document 1.

The physical CPU core 101 included in the CPU resource 100 is used as a computer resource allocated to the virtual computer 140 generated by the virtualization management unit 120 such as a hypervisor. In the following description, the virtual machine is also referred to as VM.

In the first embodiment, the virtualization management unit 120 described later converts a plurality of physical CPU cores 101 included in the CPU resource 100 into two groups of a first physical CPU core group 104 and a second physical CPU core group 105. Manage separately.

The first physical CPU core group 104 is a group of physical CPU cores 101 assigned to the virtual CPU core 150 of the VM 140 included in the priority group 190. As will be described later, the virtualization management unit 120 controls the physical CPU cores 101 included in the first physical CPU core group 104 so that transition to the root mode does not occur. As a result, it is possible to suppress a decrease in processing performance of the guest OS 160 running on the VM 140 included in the priority group 190.

On the other hand, the second physical CPU core group 105 is a group of physical CPU cores 101 other than the physical CPU core 101 included in the first physical CPU core group 104. The physical CPU cores 101 included in the second physical CPU core group 105 are allowed to transition to the root mode.

The memory resource 110 includes a plurality of memories 111. In the first embodiment, a first memory group 112 and a second memory group 113 are configured using a plurality of memories 111. The first memory group 112 is used to allocate a memory area to the virtualization management unit 120. The second memory group 113 is used to allocate a memory area to each of the plurality of VMs 140. Note that the virtualization management unit 120 can access a part or all of the memory areas of the second memory group 113.

The interrupt device resource 115 includes one or more interrupt devices 116 that generate an interrupt to the physical CPU core 101. The interrupt device 116 may be, for example, a PCIe device for I / O processing, a timer resource associated with the physical CPU core 101, or the like.

(Software configuration)
Next, the software configuration of the physical computer 10 will be described with reference to FIG.

In the physical computer 10, the virtualization management unit 120 operates as system software. The virtualization management unit 120 generates one or more VMs 140 using the computer resources of the physical computer 10 and manages the generated VMs 140. Specifically, the virtualization management unit 120 assigns one or more physical CPU cores 101 included in the CPU resource 100 to the virtual CPU core 150 and allocates a partial memory area of the second memory group 113 to the virtual memory 151. In addition, the VM 140 is generated by assigning one or more interrupt devices 116 included in the interrupt device resource 115 to the virtual interrupt device 152.

The virtualization management unit 120 according to the first embodiment includes an interrupt handler 137, a timer scheduler 138, and a setting UI 139, and also includes a physical vector number management table 130, a CPU interrupt identifier management table 131, and a VM interrupt device management table. 132, a VM attribute table 133, a physical CPU core group management table 134, high-speed IPI control data 135, and a VMCS (Virtual Machine Control Structure) 136 are held.

The interrupt handler 137 is an execution code of processing for an interrupt received by the physical CPU core 101 included in the second physical CPU core group 105. The timer scheduler 138 manages timer interrupts. The setting UI 139 is an interface for performing various settings of the VM 140. In the first embodiment, the user sets the attribute value of the guest OS processing priority mode using the setting UI 139.

The physical vector number management table 130 stores information for managing interrupts of the interrupt device 116 included in the interrupt device resource 115. Details of the physical vector number management table 130 will be described later with reference to FIG. The CPU interrupt identifier management table 131 stores information for managing the correspondence between the virtual CPU core 150 included in each VM 140 and the physical CPU core 101 included in the CPU resource 100. Details of the CPU interrupt identifier management table 131 will be described later with reference to FIG.

The VM interrupt device management table 132 stores information for managing the correspondence between physical vector numbers and virtual vector numbers managed using a virtual vector number management table 171 described later. The correspondence relationship between a physical vector number and a virtual vector number includes, for example, a correspondence relationship in which one physical vector number is occupied by one virtual vector number, and a correspondence relationship in which one physical vector number is shared by a plurality of virtual vector numbers. Conceivable. Details of the VM interrupt device management table 132 will be described later with reference to FIG.

The VM attribute table 133 stores information for managing the control mode of the interrupt process of the guest OS 160 operating on each VM 140. Details of the VM attribute table 133 will be described later with reference to FIG. The physical CPU core group management table 134 stores information for managing a group to which each physical CPU core 101 included in the CPU resource 100 belongs. Details of the physical CPU core group management table 134 will be described later with reference to FIG.

The high-speed IPI control data 135 stores information necessary for using the interrupt virtualization support function unit 102 included in each physical CPU core 101 included in the CPU resource 100. Details of the high-speed IPI control data 135 will be described later with reference to FIG.

The VMCS 136 stores information for managing the execution state of the virtual CPU core 150 included in the VM 140. Since VMCS 136 is described in Non-Patent Document 1, detailed description thereof is omitted.

The VM 140 includes a virtual CPU core 150, a virtual memory 151, and a virtual interrupt device 152. One of the physical CPU cores 101 included in the CPU resource 100 is allocated to the virtual CPU core 150, the memory area of the second memory group 113 is allocated to the virtual memory 151, and the virtual interrupt device 152 Is assigned one of the interrupt devices 116 included in the interrupt device resource 115.

The virtualization management unit 120 according to the first embodiment manages a plurality of VMs 140 divided into a priority group 190 and a non-priority group 191 according to an interrupt processing operation policy (control mode of interrupt processing).

In the VM 140 included in the priority group 190, the attribute value indicating the guest OS processing priority mode is validated. On the other hand, the VM 140 included in the non-priority group 191 has an invalid attribute value indicating the guest OS processing priority mode. Here, the guest OS processing priority mode indicates a control mode of interrupt processing of the guest OS 160 operating on the VM 140.

In the following description, the VM 140 in which the attribute value indicating the guest OS processing priority mode is validated, that is, the VM 140 included in the priority group 190 is also described as the priority mode VM 140. Further, the VM 140 in which the attribute value indicating the guest OS processing priority mode is invalidated, that is, the VM 140 included in the non-priority group 191 is also referred to as a non-priority mode VM 140.

The virtualization management unit 120 manages the VM 140 in the priority mode so as to avoid a decrease in the processing performance of the guest OS 160 due to the HW interrupt 180. On the other hand, the virtualization management unit 120 manages the VM 140 in the non-priority mode so as to allow a decrease in the processing performance of the guest OS 160 due to the HW interrupt 180.

Specifically, the physical CPU core 101 allocated to the virtual CPU core 150 included in the VM 140 in the priority mode is controlled so that the transition to the root mode does not occur. The physical CPU core 101 allocated to the virtual CPU core 150 included in the VM 140 in the non-priority mode is controlled so as to be able to transition to the root mode as in the prior art, and receives the HW interrupt 180 to the VM 140 in the priority mode. The physical CPU core 101 is also handled.

The guest OS 160 is software that operates on the VM 140 and holds a device driver 170 and a virtual vector number management table 171.

The device driver 170 is software for controlling the virtual interrupt device 152. The guest OS 160 performs an interrupt setting 184 for the virtualization management unit 120 using the device driver 170.

The virtual vector number management table 171 stores information for the guest OS 160 to manage the interruption of the virtual interrupt device 152. The virtual vector number management table 171 is updated in response to the interrupt setting 184. Details of the virtual vector number management table 171 will be described later with reference to FIG.

When the guest OS 160 is activated, an interrupt scheduling setting 185 for setting an HW interrupt from the virtual interrupt device 152 of the VM 140 to the virtual CPU core 150 is generated in the virtualization management unit 120.

Here, an overview of interrupt processing in the VM 140 in the priority mode in the first embodiment will be described.

In the HW interrupt from the virtual interrupt device 152 to the virtual CPU core 150 of the VM 140 in the priority mode, the virtualization management unit 120 determines in advance the interrupt destination of the interrupt device 116 assigned to the virtual interrupt device 152. The physical CPU core 101 is set to the physical CPU core 101 included in the second physical CPU core group 105.

First, the physical CPU core 101 included in the second physical CPU core group 105 receives the HW interrupt 180 from the interrupt device 116 assigned to the VM 140 in the priority mode. The physical CPU core 101 that has received the HW interrupt 180 makes a transition from the non-root mode to the root mode, and issues a physical IPI 181 to the physical CPU core 101 assigned to the virtual CPU core 150 of the VM 140 in the priority mode.

The received physical CPU core 101 executes the reference processing 182 of the high-speed IPI control data 135 and issues a virtual interrupt 183 directly to the virtual CPU core 150 without going through the virtualization management unit 120 based on the processing result. .

The physical CPU core 101 allocated to the virtual CPU core 150 of the priority mode VM 140 by the above-described processing can issue the virtual interrupt 183 to the priority mode VM 140 without performing the transition to the root mode. Can continue. Therefore, it is possible to suppress a decrease in processing performance of the guest OS 160 on the VM 140 in the priority mode.

On the other hand, for the interrupt processing of the VM 140 belonging to the non-priority group 191, the following processing is executed. The second physical CPU core group 105 receives the HW interrupt 180 from the interrupt device 116 and issues a virtual interrupt 186 accompanied by arbitration by the virtualization management unit 120 to the virtual CPU core 150.

In the first embodiment, the virtualization management unit 120 receives the virtual interrupt 186 for both the HW interrupt 180 for the VM 140 in the priority mode and the VM 140 in the non-priority mode, executes the arbitration process, and then performs a virtual interrupt to the VM 140. Can provide a secure virtualization environment.

Next, details of various information will be described with reference to FIGS.

FIG. 2 is an explanatory diagram illustrating an example of a physical vector number management table 130 held by the virtualization management unit 120 according to the first embodiment.

The physical vector number management table 130 includes one record for one physical vector number, and each record includes a physical vector number 210 and an interrupt device type 220.

The physical vector number 210 is a vector number (physical vector number) that can be managed by the physical computer 10. The physical vector number is information included in the HW interrupt 180. The interrupt device type 220 is a type of the interrupt device 116 registered for the physical vector number 210. When the interrupt device type 220 is blank, it indicates that no interrupt device 116 is registered for the vector number.

FIG. 3 is an explanatory diagram illustrating an example of the virtual vector number management table 171 held by the VM 140 according to the first embodiment.

The virtual vector number management table 171 includes one record for one virtual vector number, and each record includes a virtual vector number 310 and an interrupt device type 320.

The virtual vector number 310 is a vector number (virtual vector number) that can be managed by the VM 140. The virtual vector number is information included in the virtual interrupts 183 and 186. The interrupt device type 320 is a type of the virtual interrupt device 152 registered for the virtual vector number 310.

FIG. 4 is an explanatory diagram illustrating an example of the CPU interrupt identifier management table 131 held by the virtualization management unit 120 according to the first embodiment.

The CPU interrupt identifier management table 131 is a part of CPU scheduling information specialized for interrupt processing managed by the virtualization management unit 120. The CPU interrupt identifier management table 131 includes one record for one VM 140. Each record includes a VM number 410, a virtual CPU core LAPIC ID 420, and a physical CPU core LAPIC ID 430.

The VM number 410 is an identification number for uniquely identifying the VM 140. The virtual CPU core LAPIC ID 420 is an identification number of a LAPIC (Local Advanced Programmable Interrupt Controller) of the virtual CPU core 150 of the VM 140. The physical CPU core LAPIC ID 430 is an identification number of the LAPIC of the physical CPU core 101 assigned to the virtual CPU core 150.

FIG. 5 is an explanatory diagram illustrating an example of the VM interrupt device management table 132 held by the virtualization management unit 120 according to the first embodiment.

The VM interrupt device management table 132 is a part of mapping information of physical vector numbers and virtual vector numbers assigned to the interrupt devices 116 and virtual interrupt devices 152 having the same interrupt device type. The VM interrupt device management table 132 includes one record for one physical vector number. Each record has a physical vector number 510, a VM number 520, a virtual vector number 530, a virtual CPU core LAPIC ID 540, and an interrupt reception physical. Includes CPU core LAPIC ID550.

The physical vector number 510 is the same as the physical vector number 210. The VM number 520 is an identification number of the VM 140 to which the physical vector number 210 is assigned, and is the same as the VM number 410. The virtual vector number 530 is a virtual vector number mapped to the physical vector number, and is the same as the virtual vector number 310.

The virtual CPU core LAPIC ID 540 is the LAPIC identification number of the virtual CPU core 150 that is the notification destination of the HW interrupt corresponding to the virtual vector number. The interrupt reception physical CPU core LAPIC ID 550 is an identification number of the LAPIC of the physical CPU core 101 that receives the HW interrupt to the virtual CPU core 150.

FIG. 6 is an explanatory diagram illustrating an example of the VM attribute table 133 held by the virtualization management unit 120 according to the first embodiment.

The VM attribute table 133 is information for managing a group of VMs 140 corresponding to the interrupt processing control mode. The VM attribute table 133 includes one record for one VM 140, and each record includes a VM number 610 and a guest OS processing priority mode 620.

The VM number 610 is the same as the VM number 410. The guest OS processing priority mode 620 is an attribute value of the interrupt processing control mode in the guest OS 160 on each VM 140, that is, the guest OS processing priority mode. The guest OS processing priority mode 620 stores either “invalid” or “valid”. When the guest OS processing priority mode 620 is “invalid”, this indicates that the guest OS processing priority mode is invalid, that is, the VM 140 included in the non-priority group 191. On the other hand, when the guest OS processing priority mode 620 is “valid”, this indicates that the guest OS processing priority mode is valid, that is, the VM 140 included in the priority group 190.

FIG. 7 is an explanatory diagram illustrating an example of the physical CPU core group management table 134 held by the virtualization management unit 120 according to the first embodiment.

The physical CPU core group management table 134 is information for managing a group of physical CPU cores 101. The physical CPU core group management table 134 includes one record for the LAPIC identification number of one physical CPU core 101, and each record includes a physical CPU core LAPIC ID 710 and a physical CPU core group 720.

The physical CPU core LAPIC ID 710 is the same as the physical CPU core LAPIC ID 330. The physical CPU core group 720 is information for identifying a group to which the physical CPU core 101 belongs. In the first embodiment, the physical CPU core group 720 stores “1” indicating the first physical CPU core group 104 and “2” indicating the second physical CPU core group 105. Note that “2” is set in the physical CPU core group 720 in the initial state.

FIG. 8 is an explanatory diagram illustrating an example of the high-speed IPI control data 135 held by the virtualization management unit 120 according to the first embodiment.

The high-speed IPI control data 135 includes an information reading area 800 and an information writing area 810.

The information reading area 800 includes a table composed of a VM number 821, a physical CPU core LAPIC ID 822, a notification vector (Notification Vector) number 823, and a virtual vector number 824, a physical CPU core LAPIC ID 831 and a PID (Posted Interrupt Descriptor) address 832. Contains a table consisting of The notification vector number 823 and the PID address 832 are values included in the VMCS 136. The physical IPI 181 includes a notification vector number.

The information writing area 810 includes a plurality of PID data 840. The PID data 840 includes an ON (Outstanding Notification) bit 841 and a PIR (Posted Interrupt Requests) bitmap 842.

Next, processing in the physical computer 10 according to the first embodiment will be described with reference to FIGS.

(Process when the user sets the control mode of VM interrupt processing)
FIG. 9 is a flowchart illustrating an example of setting processing of the VM attribute table 133 executed by the virtualization management unit 120 according to the first embodiment.

When the virtualization management unit 120 receives the setting information of the control mode of the interrupt processing of the VM 140 from the user via the setting UI 139, the virtualization management unit 120 starts processing described below.

The virtualization management unit 120 acquires the identification number of the VM 140 and the attribute value of the guest OS processing priority mode (control mode of the interrupt processing of the VM 140) included in the setting information input via the setting UI 139 (Step S901). .

The virtualization management unit 120 generates the VM attribute table 133 based on the information acquired in step S901 (step S902). For example, the virtualization management unit 120 generates records as many as the number of VMs 140 generated on the physical computer 10, sets the identification number of the VM 140 to the VM number 610 of each generated record, and the guest of each record An attribute value is set in the OS processing priority mode 620.

When the VM attribute table 133 has already been generated, the virtualization management unit 120 updates the record of the VM attribute table 133 based on the acquired information.

(Processes executed at the start of the startup process and termination process of the VM 140)
FIG. 10 is a flowchart illustrating an example of an update process of the physical CPU core group management table 134 executed by the virtualization management unit 120 according to the first embodiment.

The virtualization management unit 120 starts processing described below at the start of the VM 140 activation processing or termination processing. The virtualization management unit 120 manages the configuration of the virtual CPU core 150 of each VM 140 and the correspondence relationship between the physical CPU cores 101 allocated to each virtual CPU core 150 at the time of starting the processing described below. Is held.

The virtualization manager 120 searches the VM 140 in the priority mode with reference to the VM attribute table 133 (step S1001), and determines whether the VM 140 in the priority mode exists (step S1002).

Specifically, the virtualization management unit 120 searches for a record in which the value of the guest OS processing priority mode 620 is “valid”. When one or more records having a value of “valid” in the guest OS processing priority mode 620 are searched, the virtualization management unit 120 determines that the VM 140 in the priority mode exists.

The virtualization management unit 120 refers to the CPU interrupt identifier management table 131 and identifies the physical CPU core 101 used by the VM 140 for each VM 140 in the priority mode (step S1003). Specifically, the following processing is executed.

The virtualization management unit 120 selects one VM 140 record to be processed when a plurality of priority mode VM 140 records are searched in step S1001. For example, the virtualization management unit 120 selects the records of the VM 140 in ascending order of the VM number 610.

The virtualization management unit 120 searches for a record that matches the VM number 610 of the record for which the VM number 410 is selected. The virtualization management unit 120 identifies the physical CPU core 101 used by the VM 140 corresponding to the selected record, based on the physical CPU core LAPIC ID 430 of the retrieved record.

The virtualization management unit 120 repeatedly executes the process described above for all the VM 140 records searched in step S1001. The above is the description of the process in step S1003.

The virtualization management unit 120 updates the physical CPU core group management table 134 (step S1004) and ends the process. The processing in step S1004 differs depending on the processing trigger.

When the activation process of the VM 140 is a processing trigger, the virtualization management unit 120 searches for a record in which the physical CPU core LAPIC ID 710 matches the physical CPU core LAPIC ID 430 of the record searched in step S1003. The virtualization management unit 120 updates the value of the physical CPU core group 720 of the retrieved record to “1”.

When the VM 140 end process is a trigger for processing, the virtualization management unit 120 searches for a record from the physical CPU core group management table 134 by the same processing. Furthermore, the virtualization management unit 120 updates the value of the physical CPU core group 720 of the retrieved record to “2”.

(Processing of the virtualization management unit when setting the guest OS interrupt)
FIG. 11 is a flowchart illustrating an example of an interrupt setting process of the guest OS 160 executed by the virtualization management unit 120 according to the first embodiment.

When the virtualization management unit 120 detects the interrupt scheduling setting 185 from the guest OS 160 or the interrupt setting 184 of the device driver 170, the virtualization management unit 120 starts processing described below.

When the virtualization management unit 120 detects the interrupt setting 184 from the device driver 170 or the interrupt scheduling setting 185 from the guest OS 160, the virtualization management unit 120 traps the interrupt setting 184 or the interrupt scheduling setting 185 (step S1101). At this time, the virtualization management unit 120 determines the identification number of the target VM 140, the target virtual vector number, the interrupt device type, and the virtual CPU core of the interrupt destination from the trapped interrupt setting 184 or interrupt scheduling setting 185. The identification number of 150 LAPIC is acquired.

The virtualization management unit 120 refers to the VM attribute table 133 and determines whether or not the target VM 140 is the VM 140 in the priority mode (step S1102).

Specifically, the virtualization management unit 120 searches for a record in which the VM number 610 matches the identification number of the target VM 140 acquired in step S1101, and the guest OS processing priority mode 620 of the searched record is “valid. Is determined. When the guest OS processing priority mode 620 of the retrieved record is “valid”, the virtualization management unit 120 determines that the target VM 140 is the VM 140 in the priority mode.

When it is determined that the target VM 140 is not the VM 140 in the priority mode, the virtualization management unit 120 ends the process.

When it is determined that the target VM 140 is the VM 140 in the priority mode, the virtualization management unit 120 refers to the physical vector number management table 130 and acquires a physical vector number (step S1103).

Specifically, the virtualization management unit 120 searches for a record in which the interrupt device type 220 matches the interrupt device type acquired in step S1101, and obtains a physical vector number from the physical vector number 210 of the searched record. get.

The virtualization management unit 120 specifies the combination of the identification number of the target VM 140, the virtual vector number, the LAPIC identification number of the interrupted virtual CPU core 150, and the physical vector number as a VM interrupt device description record (step S1104).

The virtualization manager 120 determines whether the interrupt mask setting of the trapped interrupt setting 184 or the interrupt scheduling setting 185 is “Yes” (step S1105). For example, the virtualization management unit 120 makes a determination based on preset interrupt mask setting information.

If the interrupt mask setting is “none”, the virtualization management unit 120 adds a record to the VM interrupt device management table 132 based on the VM interrupt device description record specified in step S1104 (step S1104). S1106).

Specifically, the virtualization management unit 120 adds one record to the VM interrupt device management table 132, and adds the physical vector number 510, VM number 520, virtual vector number 530, and virtual CPU core of the added record. The value included in the VM interrupt device description record is set for each LAPIC ID 540. At this time, the interrupt reception physical CPU core LAPIC ID 550 of the record added to the VM interrupt device management table 132 is blank.

The virtualization manager 120 adds information to the high-speed IPI control data 135 and updates the VMCS 136 of the target VM 140 (step S1107). Specifically, the following processing is executed.

The virtualization management unit 120 refers to the CPU interrupt identifier management table 131 and searches for a record in which the VM number 410 matches the identification number of the VM 140 in the VM interrupt device description record. The CPU core LAPIC ID 420 is searched for a line that matches the LAPIC identification number of the virtual CPU core 150 in the VM interrupt device description record. The virtualization management unit 120 acquires the value of the physical CPU core LAPIC ID 430 of the searched row.

The virtualization manager 120 adds a record to the upper table included in the information reading area 800 of the high-speed IPI control data 135. Further, the virtualization management unit 120 sets the value included in the VM interrupt device description record in the VM number 821 and the virtual vector number 824 of the added record, and the physical CPU core LAPIC ID 822 has a physical CPU. Set the value obtained from the core LAPIC ID430. In addition, the virtualization management unit 120 sets an arbitrary unique value indicating that transfer is necessary for the notification vector number 823.

Also, the virtualization management unit 120 searches for a record in which the physical CPU core LAPIC ID 831 matches the value acquired from the physical CPU core LAPIC ID 430 and acquires the address from the PID address 832 of the target physical CPU core 101. The virtualization manager 120 sets the acquired address and the value of the notification vector number 823 of the record added to the high-speed IPI control data 135 in the VMCS 136. The above is the description of the process in step S1107.

The virtualization management unit 120 registers the interrupt handler 137 for any physical CPU core 101 included in the second physical CPU core group 105, and updates the VM interrupt device management table 132 (step S1108). ). Thereafter, the virtualization management unit 120 ends the process. Specifically, the following processing is executed.

The virtualization management unit 120 refers to the physical CPU core group management table 134 and selects one record whose physical CPU core group 720 is “2”. That is, the physical CPU core 101 that receives an interrupt from the interrupt device 116 assigned to the VM 140 in the priority mode is selected from the physical CPU cores 101 included in the second physical CPU core group 105.

There are various methods for selecting the physical CPU core 101 described above. For example, the virtualization management unit 120 preferentially selects a physical CPU core LAPIC ID 710 having a small value. Also, the virtualization management unit 120 refers to the CPU interrupt identifier management table 131 and selects a physical CPU core 101 that is not assigned to the VM 140. The selection method described above is an example, and the present invention is not limited to this.

The virtualization manager 120 registers the interrupt handler 137 in the interrupt handler register 103 of the physical CPU core 101 corresponding to the physical CPU core LAPIC ID 710 of the selected record. The interrupt handler 137 is an execution code that implements a series of processes described in FIG. Further, the virtualization management unit 120 sets the value of the physical CPU core LAPIC ID 710 of the selected record in the interrupt reception physical CPU core LAPIC ID 550 of the record added to the VM interrupt device management table 132 in step S1106. .

In the first embodiment, the virtualization management unit 120 selects one physical CPU core 101 that receives the HW interrupt 180, but two or more physical CPU cores 101 may be selected. In this case, the LAPIC identification numbers of the plurality of physical CPU cores 101 are set in the interrupt reception physical CPU core LAPIC ID 550. The above is the description of the processing in step S1108.

If the interrupt mask setting is “Yes” in step S1105, the virtualization management unit 120 deletes the record from the VM interrupt device management table 132 based on the VM interrupt device description record specified in step S1104. (Step S1109).

Specifically, the virtualization management unit 120 displays a record in which each value of the physical vector number 510, the VM number 520, the virtual vector number 530, and the virtual CPU core LAPIC ID 540 matches the value of the VM interrupt device description record. Search and delete searched records. At this time, the virtualization management unit 120 temporarily holds the value of the interrupt reception physical CPU core LAPIC ID 550 of the retrieved record.

The virtualization management unit 120 deletes information from the high-speed IPI control data 135, and updates the VMCS 136 of the target VM 140 (step S1110).

Specifically, the virtualization management unit 120 records or records that the VM number 821 and the virtual vector number 824 of the information read area 800 match the identification number and virtual vector number of the VM 140 included in the VM interrupt device description record. Is deleted. In addition, the virtualization management unit 120 deletes the value of the PID address 832 from the VMCS 136.

The virtualization management unit 120 deletes the interrupt handler 137 for receiving the HW interrupt 180 for the VM 140 in the priority mode (step S1111). Thereafter, the virtualization management unit 120 ends the process.

Specifically, the virtualization management unit 120 selects the physical CPU core 101 that receives the HW interrupt 180 for the VM 140 in the priority mode based on the value of the interrupt reception physical CPU core LAPIC ID 550 of the record deleted in step S1109. Identify. The virtualization manager 120 deletes the interrupt handler 137 from the interrupt handler register 103 of the specified physical CPU core 101.

In this process, the virtual CPU core LAPIC ID 420 in the CPU interrupt identifier management table 131 is also updated, but is omitted because it is a general process. Further, the processing order from step S1106 to step S1108 may be changed so that consistency is maintained, and the processing content may be changed. For example, in step S1106, the virtualization management unit 120 may add a record to the VM interrupt device management table 132 after selecting the physical CPU core 101 that accepts the HW interrupt 180.

(Processing during virtual CPU core scheduling by the virtualization manager)
FIG. 12 is a flowchart for describing processing executed by the virtualization management unit 120 according to the first embodiment during virtual CPU core scheduling.

Here, the virtual CPU core scheduling indicates a process of changing the physical CPU core 101 assigned to the virtual CPU core 150 of each VM 140. The process described below is executed after virtual CPU core scheduling. At the start of the process, the identification number of the VM 140 to be scheduled, the identification number of the LAPIC of the virtual CPU core 150 to be scheduled, and the physical CPU core 101 assigned to the virtual CPU core 150 to be scheduled before virtual CPU core scheduling The processing result obtained by combining the identification number of the LAPIC and the identification number of the LAPIC of the physical CPU core 101 assigned to the virtual CPU core 150 to be scheduled after virtual CPU core scheduling is input.

In the following description, the physical CPU core 101 assigned to the virtual CPU core 150 before the virtual CPU core scheduling is also referred to as a released physical CPU core 101, and is assigned to the virtual CPU core 150 after the virtual CPU core scheduling. The physical CPU core 101 is also referred to as an assigned physical CPU core 101.

When a plurality of processing results are input, the virtualization management unit 120 selects one target processing result, and executes the processing from step S1201 to step S1207 on the selected processing result. Further, when the processing from step S1201 to step S1207 is completed for the selected processing result, the virtualization management unit 120 selects another processing result and executes the same processing.

The virtualization management unit 120 refers to the VM attribute table 133 (step S1201), and determines whether the scheduling target VM 140 is the VM 140 in the priority mode (step S1202).

Specifically, the virtualization management unit 120 searches for a record in which the VM number 610 matches the identification number of the VM 140 to be scheduled, and whether or not the guest OS processing priority mode 620 of the searched record is “valid”. Determine whether. When the guest OS processing priority mode 620 is “valid”, it is determined that the scheduling target VM 140 is the VM 140 in the priority mode.

When it is determined that the scheduling target VM 140 is the VM 140 in the priority mode, the virtualization management unit 120 identifies the release physical CPU core 101 (step S1203). The virtualization manager 120 changes the group of CPU cores to which the identified release physical CPU core 101 belongs from the first physical CPU core group 104 to the second physical CPU core group 105 (step S1204).

Specifically, the virtualization management unit 120 specifies the LAPIC identification number of the release physical CPU core 101 based on the processing result. The virtualization management unit 120 refers to the physical CPU core group management table 134 and searches for a record that matches the value of the LAPIC ID of the release physical CPU core 101 in which the physical CPU core LAPIC ID 710 is specified. The virtualization manager 120 sets “2” to the physical CPU core group 720 of the retrieved record.

The virtualization management unit 120 sets the LAPIC identification number of the allocated physical CPU core 101 in the high-speed IPI control data 135 and updates the VMCS 136 (step S1205). Thereafter, the virtualization management unit 120 ends the process. Specifically, the following processing is executed.

The virtualization management unit 120 searches for a record in which the VM number 821 matches the identification number of the VM 140 to be scheduled. Further, the virtualization management unit 120 searches for a record in which the physical CPU core LAPIC ID 822 of the retrieved record matches the LAPIC identification number of the release physical CPU core 101. The virtualization manager 120 overwrites the physical CPU core LAPIC ID 822 of the retrieved row with the LAPIC identification number of the assigned physical CPU core 101.

Also, the virtualization management unit 120 searches for a record in which the physical CPU core LAPIC ID 831 matches the LAPIC of the assigned physical CPU core 101, and sets the PID address 832 of the searched record in the VMCS 136. The above is the description of the processing in step S1205.

If it is determined in step S1202 that the target VM 140 is not the VM 140 in the priority mode, the virtualization management unit 120 determines whether there is a record of the released physical CPU core 101 in the VM interrupt device management table 132. (Step S1206).

Specifically, the virtualization management unit 120 refers to the VM interrupt device management table 132 and searches for a record in which the interrupt reception physical CPU core LAPIC ID 550 matches the identification number of the LAPIC of the release physical CPU core 101.

When it is determined that the record of the released physical CPU core 101 exists in the VM interrupt device management table 132, the virtualization management unit 120 ends the process.

When it is determined that there is a record of the released physical CPU core 101 in the VM interrupt device management table 132, the virtualization management unit 120 sets the interrupt handler 137 for the released physical CPU core 101 and the assigned physical CPU core 101. Also, the VM interrupt device management table 132 is updated (step S1207). Thereafter, the virtualization management unit 120 ends the process. Specifically, the following processing is executed.

The virtualization manager 120 deletes the interrupt handler 137 from the interrupt handler register 103 of the release physical CPU core 101. Also, the virtualization management unit 120 registers the deleted interrupt handler 137 in the interrupt handler register 103 of the assigned physical CPU core 101.

The virtualization management unit 120 deletes the LAPIC identification number of the release physical CPU core 101 from the interrupt reception physical CPU core LAPIC ID 550 of the record whose VM number 520 matches the identification number of the target VM 140, and allocates the physical CPU core 101 The identification number of LAPIC is set.

Further, the virtualization management unit 120 also updates the physical CPU core LAPIC ID 430 of the CPU interrupt identifier management table 131 based on the processing result of the virtual CPU core scheduling. Note that the update of the CPU interrupt identifier management table 131 is a known technique, and thus the description thereof is omitted.

(Processing of the virtualization manager when a HW interrupt occurs)
FIG. 13 is a flowchart illustrating the interrupt arbitration process executed by the virtualization management unit 120 according to the first embodiment.

In the first embodiment, the physical CPU core 101 belonging to the second physical CPU core group 105 receives the HW interrupt 180 from each interrupt device 116. At this time, the physical CPU core 101 transitions from the non-root mode to the root mode, and the virtualization management unit 120 executed by the physical CPU core 101 starts the following processing.

The virtualization manager 120 analyzes the received HW interrupt 180 and determines whether or not the HW interrupt 180 needs to be transferred based on the analysis result (step S1301). That is, it is determined whether it is necessary to transfer the HW interrupt 180 to another physical CPU core 101.

For example, when the received HW interrupt 180 is an MSI-X interrupt based on PCIe, the virtualization management unit 120 refers to the VM interrupt device management table 132 and receives the MSI-X interrupt received by the physical vector number 510. It is determined whether there is a record that matches the physical vector number included in the file. If there is a record that satisfies the above-described condition in the VM interrupt device management table 132, the virtualization management unit 120 determines that the received HW interrupt 180 needs to be transferred.

When the received HW interrupt 180 is a timer interrupt of the physical CPU core 101, the virtualization management unit 120 records in the VM interrupt device management table 132 a record that matches the physical vector number included in the timer interrupt. It is determined whether or not exists. If there is a record that satisfies the above-described condition, the virtualization management unit 120 further determines whether or not the VM 140 that the timer scheduler 138 determines to issue an interrupt next is the VM 140 in the priority mode. When the VM 140 is the VM 140 in the priority mode, the virtualization management unit 120 determines that the received HW interrupt 180 needs to be transferred.

When it is determined that it is necessary to transfer the received HW interrupt 180, the virtualization management unit 120 refers to the VM interrupt device management table 132 and obtains information on the VM 140 that is the transfer destination of the received HW interrupt 180. Obtain (step S1302).

Specifically, the virtualization management unit 120 searches for a record in which the physical vector number 510 matches the physical vector number included in the received HW interrupt 180, and the VM number 520 and virtual vector number 530 of the searched record. And the value of the virtual CPU core LAPIC ID 540 is acquired.

The virtualization management unit 120 refers to the CPU interrupt identifier management table 131 and identifies the physical CPU core 101 to which the received HW interrupt 180 is transferred (step S1303).

Specifically, the virtualization management unit 120 determines that the VM number 410 matches the value acquired from the VM number 520, and the virtual CPU core LAPIC ID 420 matches the value acquired from the virtual CPU core LAPIC ID 540. Search for. The virtualization management unit 120 acquires the value of the physical CPU core LAPIC ID 430 of the searched row.

The virtualization manager 120 updates the PID data 840 of the high-speed IPI control data 135 in order to set the virtual vector number in the physical IPI 181 (step S1304). Specifically, the following processing is executed.

The virtualization management unit 120 searches for a record in which the physical CPU core LAPIC ID 831 matches the value of the physical CPU core LAPIC ID 430. The virtualization management unit 120 reads the PID data 840 based on the value of the PID address 832 of the retrieved record.

The virtualization management unit 120 validates the bit corresponding to the value acquired from the virtual vector number 530 in step S1302 among the bits included in the PIR bitmap 842 of the PID data 840. For example, when the value of the virtual vector number 530 is “x”, the virtualization management unit 120 updates the value of the “x” -th bit in the PIR bitmap 842 to “1”.

Further, the virtualization management unit 120 determines whether or not the ON bit 841 is enabled. When the ON bit 841 is not enabled, the virtualization management unit 120 enables the ON bit 841. For example, the virtualization management unit 120 determines whether the ON bit 841 is “1”, and updates the value of the bit to “1” when the ON bit 841 is “0”. The above is the description of the processing in step S1304.

The virtualization management unit 120 issues a physical IPI 181 to the physical CPU core 101 assigned to the virtual CPU core 150 of the VM 140 in the priority mode (step S1305). Thereafter, the virtualization management unit 120 ends the process. Specifically, the following processing is executed.

The virtualization management unit 120 refers to the information reading area 800, the VM number 821, the virtual vector number 824, and the physical CPU core LAPIC ID 822 are the virtual vector number 530, the virtual CPU core LAPIC ID 540, and the physical CPU core LAPIC ID 430. Search for records that match the value. The virtualization management unit 120 acquires the value of the notification vector number 823 of the retrieved record.

The virtualization management unit 120 issues a physical IPI 181 to which the value of the notification vector number 823 is added to the physical CPU core specified in step S1303. The above is the description of the processing in step S1305.

The physical CPU core 101 that has received the physical IPI 181 executes the following processing.

First, the physical CPU core 101 that has received the physical IPI 181 refers to the high-speed IPI control data 135 and searches for a record in which the physical CPU core LAPIC ID 831 matches the identification number of the LAPIC of the physical CPU core 101. The physical CPU core 101 accesses the PID data 840 based on the value of the PID address 832 of the retrieved record, and reads the value of the PIR bitmap 842 from the PID data 840. The physical CPU core 101 specifies a virtual vector number from the value of each bit of the read PIR bitmap 842. For example, when the value of the xth bit is “1”, the virtual vector number is “x”.

The physical CPU core 101 that has received the physical IPI 181 refers to the CPU interrupt identifier management table 131 and searches for a row in which the physical CPU core LAPIC ID 430 matches the identification number of the LAPIC of the physical CPU core 101. The physical CPU core 101 that has received the physical IPI 181 issues a virtual interrupt 183 including the identified virtual vector number to the virtual CPU core 150 corresponding to the virtual CPU core LAPIC ID 420 in the retrieved row. The above is the description of the processing of the physical CPU core 101 that has received the physical IPI 181.

If it is determined in step S1301 that it is not necessary to transfer the received HW interrupt 180, the virtualization manager 120 executed by the physical CPU core 101 that has received the HW interrupt 180, A virtual interrupt 186 is issued to the virtual CPU core 150 of the VM 140 of the priority group (step S1306). Thereafter, the virtualization management unit 120 ends the process.

(Processing when the virtualization management unit 120 is activated)
FIG. 14 is a flowchart illustrating an example of a startup process executed by the virtualization management unit 120 according to the first embodiment. In the following description, a description of processing of the general virtualization management unit 120 such as securing and initializing a storage area for storing the physical vector number management table 130 is omitted.

The virtualization management unit 120 acquires a memory area for storing the VM interrupt device management table 132, the VM attribute table 133, the physical CPU core group management table 134, and the high-speed IPI control data 135 from the first memory group 112. (Step S1401).

The virtualization management unit 120 initializes the VM interrupt device management table 132, the VM attribute table 133, the physical CPU core group management table 134, and the high-speed IPI control data 135 in the acquired memory area (step S1402). Thereafter, the virtualization management unit 120 ends the process. Specifically, the following processing is executed.

The virtualization management unit 120 generates an empty VM interrupt device management table 132. The virtualization manager 120 generates records for the identification number of the maximum number of VMs 140 that can be defined by the virtualization manager 120 in the VM attribute table 133. Further, the virtualization management unit 120 sets the VM 140 identification number in ascending order to the VM number 610 of each record, and sets “invalid” to the guest OS processing priority mode 620 of each record.

The virtualization management unit 120 generates records for the number of physical CPU cores 101 included in the CPU resource 100 in the physical CPU core group management table 134. The virtualization management unit 120 sets the LAPIC identification number of each physical CPU core 101 in the physical CPU core LAPIC ID 710 of each record, and sets “2” in the physical CPU core group 720 of each record.

The virtualization management unit 120 sets, for the high-speed IPI control data 135, each head address of the PID data 840 corresponding to the LAPIC identification numbers of all the physical CPU cores 101 included in the CPU resource 100 as the PID address 832. .

(Summary)
As described above, according to the first embodiment, the physical CPU core 101 included in the second physical CPU core group receives the HW interrupt 180 from the interrupt device resource 115 for the VM 140 in the priority mode. Can be controlled. The physical CPU core 101 that has received the HW interrupt 180 issues a physical IPI 181 to the virtual CPU core 150 assigned to the VM 140 in the priority mode, and the physical CPU core 101 that has received the physical IPI 181 The virtual interrupt 183 can be issued directly to the virtual CPU core 150 without going through 120.

Therefore, the virtualization management unit 120 is always involved in both the HW interrupt 180 of the VM 140 in the priority mode and the VM 140 in the non-priority mode. Since the HW interrupt 180 is not directly transmitted to the VM 140 without involving the virtualization management unit 120, a secure virtualization environment can be provided.

In addition, since the physical CPU core 101 allocated to the priority mode VM 140 does not transition from the non-root mode to the root mode, the processing performance of the priority mode VM 140 for the HW interrupt 180 in a secure virtualization environment Can be improved.

In the second embodiment, the physical CPU core 101 that receives the HW interrupt 180 for the VM 140 in the priority mode, that is, the physical CPU core 101 that always operates in the root mode is set in advance for the physical computer 10. Hereinafter, the second embodiment will be described focusing on differences from the first embodiment.

(Hardware configuration and software configuration)
Since the hardware configuration and software configuration of the physical computer 10 of the second embodiment are the same as those of the first embodiment, description thereof is omitted. Various information of the second embodiment is the same as that of the first embodiment except for the physical CPU core group management table 134, and the description thereof will be omitted.

In the physical CPU core group 720 of the physical CPU core group management table 134 of the second embodiment, “1”, “2”, or “occupied” is stored. “Occupied” is the physical CPU core 101 that always operates in the root mode, and is used as the physical CPU core 101 that receives the HW interrupt 180 for the VM 140 in the priority mode. The physical CPU core 101 in which the physical CPU core group 720 is “occupied” is not handled as the physical CPU core 101 assigned to the VM 140.

In the following description, the physical CPU core 101 in which the physical CPU core group 720 is “occupied” is also referred to as an interrupt-dedicated physical CPU core 101.

In the second embodiment, definition information related to the physical CPU core 101 dedicated to interrupts is set in the physical computer 10 in advance. The definition information includes, for example, the LAPIC identification number of the physical CPU core 101 set as the interrupt-only physical CPU core 101. The definition information is stored in a ROM or storage medium (not shown) of the physical computer 10.

The virtualization management unit 120 according to the second embodiment sets a predetermined physical CPU core 101 as an interrupt-dedicated physical CPU core 101 based on the definition information at startup.

(Process when the user sets the control mode of VM interrupt processing)
Since the setting process of the VM attribute table 133 executed by the virtualization management unit 120 of the second embodiment is the same as that of the first embodiment, the description thereof is omitted.

(Processes executed at the start of the startup process and termination process of the VM 140)
Since the update processing of the physical CPU core group management table 134 executed by the virtualization management unit 120 of the second embodiment is the same as that of the first embodiment, description thereof is omitted.

(Processing of the virtualization manager when setting a guest OS interrupt)
FIG. 15 is a flowchart illustrating an example of an interrupt setting process of the guest OS 160 executed by the virtualization management unit 120 according to the second embodiment.

Since the processing from step S1501 to step S1506 is the same as the processing from step S1101 to step S1106, description thereof will be omitted. Further, the processing from step S1509 to step S1511 is the same as the processing from step S1109 to step S1111, and thus the description thereof is omitted.

In step S1507, the virtualization management unit 120 adds information to the high-speed IPI control data 135 and updates the VMCS 136 of the target VM 140 (step S1507). Specifically, the following processing is executed.

The virtualization manager 120 adds a record to the upper table included in the information reading area 800 of the high-speed IPI control data 135. Further, the virtualization management unit 120 sets values included in the VM interrupt device description record in the VM number 821, the notification vector number 823, and the virtual vector number 824 of the added record, and the physical CPU core LAPIC. The value acquired from the physical CPU core LAPIC ID 430 is set in ID 822.

Also, the virtualization management unit 120 searches for a record in which the physical CPU core LAPIC ID 831 matches the value acquired from the physical CPU core LAPIC ID 430 and acquires the address from the PID address 832 of the target physical CPU core 101. The virtualization management unit 120 sets the acquired address in the VMCS 136. The above is the description of the processing in step S1507.

In step S1508, the virtualization management unit 120 registers the interrupt handler 137 for the physical CPU core 101 dedicated to interrupts, and updates the VM interrupt device management table 132 (step S1508). Thereafter, the virtualization management unit 120 ends the process. Specifically, the following processing is executed.

The virtualization management unit 120 refers to the physical CPU core group management table 134 and selects one record in which the physical CPU core group 720 is “occupied”. In the second embodiment, the interrupt-only physical CPU core 101 is selected as the physical CPU core 101 that receives the HW interrupt 180 for the VM 140 in the priority mode.

There are various methods for selecting the physical CPU core 101 described above. For example, the virtualization management unit 120 preferentially selects a physical CPU core LAPIC ID 710 having a small value. The selection method described above is an example, and the present invention is not limited to this.

The virtualization manager 120 registers the interrupt handler 137 in the interrupt handler register 103 of the physical CPU core 101 corresponding to the physical CPU core LAPIC ID 710 of the selected record. Further, the virtualization management unit 120 sets the value of the physical CPU core LAPIC ID 710 of the selected record in the interrupt reception physical CPU core LAPIC ID 550 of the record added to the VM interrupt device management table 132 in step S1506. . The above is the description of the processing in step S1508.

(Processing during virtual CPU core scheduling by the virtualization manager)
The processing executed by the virtualization management unit 120 of the second embodiment at the time of virtual CPU core scheduling is the same as that of the first embodiment, and thus the description thereof is omitted.

(Processing of the virtualization management unit 120 when a HW interrupt occurs)
The interrupt arbitration process executed by the virtualization management unit 120 according to the second embodiment is the same as that according to the first embodiment, and a description thereof will be omitted. The second embodiment is different from the first embodiment in that the physical CPU core 101 that receives the HW interrupt 180 is a physical CPU core 101 dedicated to interrupts.

(Processing when the virtualization management unit 120 is activated)
FIG. 16 is a flowchart illustrating an example of a startup process executed by the virtualization management unit 120 according to the second embodiment. In the following description, a description of processing of the general virtualization management unit 120 such as securing and initializing a storage area for storing the physical vector number management table 130 is omitted.

Since the processing of step S1601 and step S1602 is the same processing as the processing of step S1401 and step S1402, description thereof will be omitted.

The virtualization management unit 120 sets a specific physical CPU core 101 as the “occupied” physical CPU core 101 among the physical CPU cores 101 included in the CPU resource 100 (step S1603). Thereafter, the virtualization management unit 120 ends the process.

The virtualization manager 120 selects the physical CPU core 101 to be set as the “occupied” physical CPU core 101 from the physical CPU cores 101 included in the CPU resource 100. For example, a method of selecting the physical CPU core 101 based on definition information preset in the physical computer 10 can be considered. Further, the virtualization management unit 120 may receive an input from the administrator using the setting UI 139 and may select the physical CPU core 101 based on the input.

The virtualization management unit 120 refers to the physical CPU core group management table 134 and searches for a record that matches the identification number of the LAPIC of the physical CPU core 101 for which the physical CPU core LAPIC ID 710 is selected. The virtualization manager 120 sets “occupied” to the physical CPU core group 720 of the retrieved record. The above is the description of the processing in step S1603.

(Summary)
As described above, according to the second embodiment, the HW interrupt 180 from the interrupt device resource 115 for the VM 140 in the priority mode can be controlled to be received by the physical CPU core 101 dedicated to the interrupt. Therefore, the same effects as those of the first embodiment can be obtained.

Further, according to the second embodiment, since the physical CPU core 101 dedicated to the interrupt, not the physical CPU core 101 allocated to the VM 140 in the non-priority mode, receives the HW interrupt 180, the processing of the VM 140 in the non-priority mode is performed. It is also possible to prevent the delay.

In the first and second embodiments, control is performed in units of physical CPU cores 101. However, control may be performed in units of physical CPUs.

In addition, this invention is not limited to the above-mentioned Example, Various modifications are included. Further, for example, the above-described embodiments are described in detail for easy understanding of the present invention, and are not necessarily limited to those provided with all the described configurations. Further, a part of the configuration of each embodiment can be added to, deleted from, or replaced with another configuration.

In addition, each of the above-described configurations, functions, processing units, processing means, and the like may be realized by hardware by designing a part or all of them with, for example, an integrated circuit. The present invention can also be realized by software program codes that implement the functions of the embodiments. In this case, a storage medium in which the program code is recorded is provided to the computer, and a CPU included in the computer reads the program code stored in the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the program code itself and the storage medium storing it constitute the present invention. Examples of storage media for supplying such program codes include flexible disks, CD-ROMs, DVD-ROMs, hard disks, SSDs (Solid State Drives), optical disks, magneto-optical disks, CD-Rs, magnetic tapes, A non-volatile memory card, ROM, or the like is used.

Further, the program code for realizing the functions described in this embodiment can be implemented by a wide range of programs or script languages such as assembler, C / C ++, Perl, Shell, PHP, Java, and the like.

Furthermore, by distributing the program code of the software that realizes the functions of the embodiments via a network, the program code is stored in a storage means such as a hard disk or memory of a computer or a storage medium such as a CD-RW or CD-R. The CPU included in the computer may read and execute the program code stored in the storage unit or the storage medium.

In the above-described embodiments, the control lines and information lines indicate those that are considered necessary for the explanation, and do not necessarily indicate all the control lines and information lines on the product. All the components may be connected to each other.

Claims

A computer on which a plurality of virtual machines generated using logically divided computer resources operate,
The computer resource includes a plurality of physical arithmetic devices, one or more physical memories, and one or more physical interrupt devices that generate hardware interrupts to the physical arithmetic devices,
Each of the plurality of virtual machines has a plurality of virtual arithmetic devices, a virtual memory, and a virtual interrupt device,
The one or more physical memories store a program for generating a plurality of virtual machines and realizing a virtualization manager that manages the plurality of virtual machines;
On the plurality of virtual machines, an operating system that controls the virtual machines runs,
The plurality of physical arithmetic devices are:
A virtualization support unit that switches between a first operation mode for executing processing by the virtualization management unit and a second operation mode for executing processing by the operating system;
An interrupt virtualization support unit that transfers the hardware interrupt issued from the one or more physical interrupt devices between the physical arithmetic devices,
The virtualization management unit
A priority group composed of virtual machines in a priority mode which is a control mode for avoiding a decrease in the processing performance of the operating system due to the hardware interrupt, and the processing performance of the operating system due to the hardware interrupt Managing the plurality of virtual machines in a non-priority group composed of virtual machines in non-priority mode, which is a control mode that allows a decrease in
A physical computing device other than the physical computing device assigned to the priority mode virtual machine is issued from the one or more physical interrupt devices assigned to the virtual interrupt device of the priority mode virtual machine. Set as a physical computing device to receive the hardware interrupt,
The virtualization management unit executed by the physical arithmetic device that has received the hardware interrupt for the virtual machine in the priority mode,
Identify the physical computing device assigned to the virtual machine in the priority mode as the transfer destination of the received hardware interrupt;
Transferring the received hardware interrupt to the identified physical processor;
The physical processing device that has received the transferred hardware interrupt is:
Identify the virtual machine that is the target of the transferred hardware interrupt,
A computer that issues a virtual interrupt to the virtual arithmetic unit included in the identified virtual computer based on the transferred hardware interrupt.
The computer according to claim 1,
The virtualization management unit
Holds physical vector number management information for managing the physical vector number used for the hardware interrupt,
Holding virtual machine interrupt device management information for controlling the hardware interrupt for the virtual machine in the priority mode,
The operating system holds virtual vector number management information for managing a virtual vector number used for the virtual interrupt,
The virtualization management unit
Detecting an interrupt setting request in the virtual interrupt device from the first operating system;
Determining whether the virtual machine on which the first operating system runs is a virtual machine in the priority mode;
When it is determined that the virtual machine on which the first operating system is operating is the virtual machine in the priority mode, the physical vector number based on the type of the virtual interrupt device included in the interrupt setting request With reference to management information, obtain the physical vector number to be included in the hardware interrupt issued from the one or more physical interrupt devices corresponding to the virtual interrupt device,
The one or more physical interrupts assigned to the virtual interrupt device of the priority mode virtual machine among the plurality of physical operation units other than the physical operation unit assigned to the priority mode virtual machine Select one physical processor that receives the hardware interrupt issued by the device,
An identification number of a virtual machine on which the first operating system operates, an identification number of the virtual vector included in the detected interrupt setting request, and an identification of the virtual processing unit included in the virtual machine on which the first operating system operates Adding a record in which the information, the acquired physical vector number, and the identification information of the selected physical arithmetic device are associated with the virtual machine interrupt device management information,
A computer, wherein an interrupt handler for processing the hardware interrupt for the virtual machine in the priority mode is set in the selected physical arithmetic device.
The computer according to claim 2,
The virtualization management unit
Holds arithmetic unit interrupt identifier management information for managing the correspondence between the virtual arithmetic unit possessed by the virtual machine and the physical arithmetic unit assigned to the virtual arithmetic unit,
The virtualization management unit executed by the physical arithmetic device that has received the hardware interrupt issued from the one or more physical interrupt devices,
Analyzing the received hardware interrupt to determine whether the received hardware interrupt needs to be transferred based on the result of the analysis;
If it is determined that it is necessary to transfer the received hardware interrupt, the virtual machine interrupt device management information is referred to based on the physical vector number included in the received hardware interrupt, The priority mode virtual machine identification information, the virtual vector number, and the virtual computing device identification information are acquired as information on the virtual machine in the priority mode for transmitting the received hardware interrupt,
The physical arithmetic device that transfers the received hardware interrupt with reference to the arithmetic device interrupt identifier management information based on the virtual machine information of the priority mode acquired from the virtual machine interrupt device management information A computer characterized by specifying.
The computer according to claim 3, wherein
The virtualization manager holds control data used for transferring the hardware interrupt between the physical arithmetic devices,
The control data includes a bitmap for specifying the virtual vector number to be included in the virtual interrupt for the virtual machine for each of the plurality of physical arithmetic devices,
The virtualization management unit executed by the physical arithmetic device that has received the hardware interrupt for the virtual machine in the priority mode,
With reference to the control data, the bitmap corresponding to the identified physical arithmetic device is accessed,
Among the bits included in the accessed bitmap, enable the bit corresponding to the virtual vector number acquired as the virtual machine information of the priority mode,
The physical processing device that has received the transferred hardware interrupt is:
With reference to the control data, access the bitmap corresponding to the physical arithmetic unit that received the transferred hardware interrupt,
Based on the accessed bitmap, obtain the virtual vector number to be included in the virtual interrupt for the virtual machine in the priority mode,
A computer that issues the virtual interrupt including the acquired virtual vector number to the virtual machine in the priority mode based on the transferred hardware interrupt.
The computer according to claim 3, wherein
The virtualization management unit
An interface for setting whether to include each of the plurality of virtual machines in the priority group or the non-priority group;
A computer that generates group management information for managing a group to which each of the plurality of virtual computers, which is determined based on information input through the interface, belongs.
The computer according to claim 5, wherein
The virtualization management unit
Referring to the group management information, identify the physical computing device assigned to the priority mode virtual machine,
Managing the physical arithmetic unit assigned to the virtual machine in the priority mode as a first arithmetic unit group;
A computer that manages the physical arithmetic devices other than the physical arithmetic device allocated to the virtual computer in the priority mode as a second arithmetic device group.
The computer according to claim 6, wherein
The virtualization management unit
Of the plurality of physical arithmetic devices, set any physical arithmetic device as a dedicated physical arithmetic device for receiving the hardware interrupt for the virtual machine in the priority mode,
Assigning the plurality of physical arithmetic units other than the dedicated physical arithmetic unit to the virtual machine,
A physical operation that receives the hardware interrupt issued from the one or more physical interrupt devices assigned to the virtual interrupt device of the priority mode virtual machine, the physical operation device dedicated to the interrupt. A computer characterized by selecting as a device.
An interrupt control method in a computer in which a plurality of virtual machines generated using logically divided computer resources are operated,
The computer resource includes a plurality of physical arithmetic devices, one or more physical memories, and one or more physical interrupt devices that generate hardware interrupts to the physical arithmetic devices,
Each of the plurality of virtual machines has a plurality of virtual arithmetic devices, a virtual memory, and a virtual interrupt device,
The one or more physical memories store a program for generating a plurality of virtual machines and realizing a virtualization manager that manages the plurality of virtual machines;
On the plurality of virtual machines, an operating system that controls the virtual machines runs,
The plurality of physical arithmetic devices are:
A virtualization support unit that switches between a first operation mode for executing processing by the virtualization management unit and a second operation mode for executing processing by the operating system;
An interrupt virtualization support unit that transfers the hardware interrupt issued from the one or more physical interrupt devices between the physical arithmetic devices,
The virtualization management unit includes a priority group composed of virtual machines in a priority mode that is a control mode for avoiding a decrease in processing performance of the operating system due to the hardware interrupt, and the hardware interrupt. Managing the plurality of virtual machines divided into non-priority groups composed of virtual machines in non-priority mode, which is a control mode that allows a reduction in processing performance of the operating system,
The interrupt control method is:
The one or more physical units assigned to the virtual interrupt device of the priority mode virtual machine, wherein the virtualization management unit has a physical calculation unit other than the physical calculation unit assigned to the priority mode virtual machine. A first step of setting as a physical arithmetic device that receives the hardware interrupt issued from the interrupt device;
The virtualization management unit executed by the physical arithmetic unit that has received the hardware interrupt for the priority mode virtual machine is assigned to the priority mode virtual machine to which the received hardware interrupt is transferred. A second step of identifying the physical arithmetic unit;
A virtualization management unit executed by the physical arithmetic device that has received the hardware interrupt for the virtual machine in the priority mode transfers the received hardware interrupt to the identified physical arithmetic device. Steps,
A fourth step in which a physical processing device that has received the transferred hardware interrupt specifies a virtual machine that is a target of the transferred hardware interrupt; and
The physical processing device that has received the transferred hardware interrupt issues a virtual interrupt to the virtual processing device of the identified virtual machine based on the transferred hardware interrupt. And an interrupt control method comprising the steps of:
The interrupt control method according to claim 8,
The virtualization management unit
Holds physical vector number management information for managing the physical vector number used for the hardware interrupt,
Holding virtual machine interrupt device management information for controlling the hardware interrupt for the virtual machine in the priority mode,
The operating system holds virtual vector number management information for managing a virtual vector number used for the virtual interrupt,
The first step includes
Detecting an interrupt setting request in the virtual interrupt device from a first operating system;
Determining whether the virtual machine running the first operating system is a virtual machine in the priority mode;
When it is determined that the virtual machine on which the first operating system is operating is the virtual machine in the priority mode, the physical vector number based on the type of the virtual interrupt device included in the interrupt setting request Referring to management information, obtaining the physical vector number to be included in the hardware interrupt issued from the one or more physical interrupt devices corresponding to the virtual interrupt device;
The one or more physical interrupts assigned to the virtual interrupt device of the priority mode virtual machine among the plurality of physical operation units other than the physical operation unit assigned to the priority mode virtual machine Selecting one physical processor that receives the hardware interrupt issued by the device;
An identification number of a virtual machine on which the first operating system operates, an identification number of the virtual vector included in the detected interrupt setting request, and an identification of the virtual processing unit included in the virtual machine on which the first operating system operates Adding a record associating information, the acquired physical vector number, and identification information of the selected physical arithmetic unit to the virtual machine interrupt device management information;
And setting an interrupt handler for processing the hardware interrupt to the virtual machine in the priority mode in the selected physical arithmetic device.
The interrupt control method according to claim 9,
The virtualization management unit
Holds arithmetic unit interrupt identifier management information for managing the correspondence between the virtual arithmetic unit possessed by the virtual machine and the physical arithmetic unit assigned to the virtual arithmetic unit,
The second step includes
The virtualization management unit executed by the physical arithmetic unit that has received the hardware interrupt issued from the one or more physical interrupt devices analyzes the received hardware interrupt, and Determining whether it is necessary to transfer the received hardware interrupt based on a result; and
It is determined that the virtualization management unit executed by the physical arithmetic unit that has received the hardware interrupt issued from the one or more physical interrupt devices needs to transfer the received hardware interrupt. The virtual machine interrupt device management information based on the physical vector number included in the received hardware interrupt, the virtual machine identification information of the priority mode, the virtual vector number, Obtaining identification information of a virtual processing device as information on the virtual machine in the priority mode for transmitting the received hardware interrupt;
The virtualization mode executed by the physical arithmetic unit that has received the hardware interrupt issued from the one or more physical interrupt devices is the priority mode acquired from the virtual machine interrupt device management information. And referring to the arithmetic device interrupt identifier management information based on the virtual computer information of the virtual machine, and specifying the physical arithmetic device to which the received hardware interrupt is transferred. Method.
The interrupt control method according to claim 10,
The virtualization manager holds control data used for transferring the hardware interrupt between the physical arithmetic devices,
The control data includes a bitmap for specifying the virtual vector number to be included in the virtual interrupt for the virtual machine for each of the plurality of physical arithmetic devices,
The second step includes
The virtualization management unit executed by the physical arithmetic unit that has received the hardware interrupt for the virtual machine in the priority mode refers to the control data, and the bitmap corresponding to the specified physical arithmetic unit Step to access
The virtualization management unit executed by the physical arithmetic unit that has received the hardware interrupt for the priority mode virtual machine has information on the priority mode virtual machine among the bits included in the accessed bitmap. Enabling a bit corresponding to the virtual vector number obtained as
The fourth step includes
Referring to the control data, accessing a bitmap corresponding to a physical processing unit that has received the transferred hardware interrupt; and
Obtaining the virtual vector number to be included in the virtual interrupt for the virtual machine in the priority mode based on the accessed bitmap;
The fifth step includes a step of issuing the virtual interrupt including the acquired virtual vector number to the virtual machine in the priority mode based on the transferred hardware interrupt. Interrupt control method.
The interrupt control method according to claim 10,
The virtualization management unit
An interface for setting whether to include each of the plurality of virtual machines in the priority group or the non-priority group;
The interrupt control method includes a step of generating group management information for managing a group to which each of the plurality of virtual machines belongs determined by the virtualization management unit based on information input through the interface. An interrupt control method characterized by the above.
The interrupt control method according to claim 12,
The virtualization management unit
Referring to the group management information, identify the physical computing device assigned to the priority mode virtual machine,
Managing the physical arithmetic unit assigned to the virtual machine in the priority mode as a first arithmetic unit group;
An interrupt control method characterized by managing the physical arithmetic devices other than the physical arithmetic devices assigned to the virtual machine in the priority mode as a second arithmetic device group.
An interrupt control method according to claim 13,
The virtualization management unit sets an arbitrary physical arithmetic device among the plurality of physical arithmetic devices as an interrupt-dedicated physical arithmetic device for receiving the hardware interrupt for the virtual machine in the priority mode. When,
The virtualization management unit includes allocating the plurality of physical arithmetic devices other than the interrupt-dedicated physical arithmetic device to the virtual machine, and
In the first step, the hardware is issued from the one or more physical interrupt devices assigned to the virtual interrupt device of the priority mode virtual machine with the interrupt dedicated physical processing unit. An interrupt control method comprising a step of selecting as a physical arithmetic device that receives an interrupt.