CN102193816B - A device allocation method and system - Google Patents

Abstract

The present invention is applicable to the technical field of computers, and provides a device allocation method and system, said method comprising the steps of: powering on the system, starting dual graphics cards in the physical BIOS, and completing the initialization operation; controlling the I/O of the virtual system BIOS O access initialization settings; intercept the DMA transmission of the device through DMA remapping hardware, and control the address conversion, and control the DMA operation of the device to directly access the memory space of the client; when the device with BIOS is allocated, by setting the client The machine shadow page table directly maps the I/O device to the MMIO space of the client, and the MMIO operation can directly access the hardware to realize efficient virtualization and allocation of the device.

Description

A device allocation method and system

technical field

The invention belongs to the technical field of computers, and in particular relates to a device allocation method and system.

Background technique

In the computer field, the virtualization of physical resources by virtual machines can be attributed to three main tasks: processor virtualization, memory virtualization, and I/O virtualization. The impact is far greater than computer-intensive, and I/O virtualization is the bottleneck restricting system performance, which will affect the stability and performance of the entire system.

In the prior art, I/O device virtualization can be classified into class virtualization, device virtualization and direct allocation virtualization.

In the quasi-virtualization system, the guest operating system does not use any existing hardware device driver, but uses a front-end and back-end interactive device driver to send I/O requests and receive I/O feedback. The end in the client operating system becomes the front-end device driver, and the end in Domain0 becomes the back-end device driver. The class virtualization device driver distinguishes the device model, and each device type uses a device driver. The front-end driver and the back-end driver communicate asynchronously through the event channel mechanism. In this way, the driver of the guest operating system is modified to make The device driver can directly interact with the hardware device through the interface provided by the virtual machine monitor, but it lacks versatility and is not conducive to promotion.

In the device virtualization system, the virtual machine monitor needs to simulate a certain target device, provide a virtual device for the client, so that it can transparently operate on the virtual device, and the guest operating system discovers the virtual target device Finally, the driver of the target device will drive the device, the driver in the client will issue some requests and wait for the response of the device, the virtual machine monitor intercepts and processes all device requests, and returns the response to the client operation The system and the original driver of the client can drive the virtual device without modification, but frequent switching and complex software design will cause performance loss of the client.

Therefore, the virtualization of I/O devices in the technical solutions including class virtualization and device virtualization provided by the prior art affects the stability and performance of the system, and has relatively large defects.

Contents of the invention

The purpose of the embodiments of the present invention is to provide a method for device allocation, which aims to solve the problem of virtualization of I/O devices in the technical solutions including class virtualization and device virtualization provided by the prior art, which affects the stability and performance of the system. Big flaw problem.

The embodiment of the present invention is achieved in this way, a device allocation method, the method includes the following steps:

Power on the system, start the dual graphics card in the physical BIOS, and complete the initialization operation;

Control the initialization settings for I/O access to the virtual system BIOS;

The DMA transmission of the device is intercepted by the DMA remapping hardware, and the address translation is controlled, and an area is allocated at the low end of the system memory to allocate it to the client system, so that the client page number and the machine page number in the client system Equal, the DMA operation of the control device directly accesses the memory space of the client;

When the device with BIOS is allocated, the I/O device is directly mapped to the MMIO space of the client by setting the client shadow page table. For devices without page alignment, control the virtual machine monitor to scan the PCI bus, and the control will The MMIO address without page alignment is changed to page alignment, and the MMIO operation can directly access hardware.

Another object of the embodiments of the present invention is to provide an equipment allocation system, the equipment allocation system is built into a computer, and the system includes:

The initialization operation mode is used to power on the system, start the dual graphics card in the physical BIOS, and complete the initialization operation;

A setting module is used to control the initial setting of I/O access to the virtual system BIOS;

The DMA remapping control module is used to intercept the DMA transmission of the device through the DMA remapping hardware, and control the address translation, and allocate a low-end area of the system memory to the client system, so that the client system The client page number is equal to the machine page number, and the DMA operation of the control device directly accesses the memory space of the client; and

The BIOS device allocation control module is used to directly map the I/O device to the MMIO space of the client by setting the shadow page table of the client when the device with the BIOS is allocated, and to control the virtual machine for the device without page alignment The monitor scans the PCI bus, controls to change the MMIO address without page alignment into page alignment, and the MMIO operation can directly access hardware.

In the embodiment of the present invention, the system is powered on, the dual graphics cards are started in the physical BIOS, and the initialization operation is completed; the initialization setting of I/O access is controlled to the virtual system BIOS; the transmission of the DMA of the device is intercepted by the DMA remapping hardware, and Control the address conversion, and control the DMA operation of the device to directly access the memory space of the client; when the device with BIOS is allocated, the I/O device is directly mapped to the MMIO space of the client by setting the client shadow page table , the MMIO operation can directly access the hardware, realizing efficient virtualization and allocation of devices.

Description of drawings

FIG. 1 is a flow chart of an implementation of a device allocation method provided by an embodiment of the present invention;

FIG. 2 is a schematic diagram of a DMA remapping mechanism provided by an embodiment of the present invention;

Fig. 3 is a schematic diagram of a device direct allocation model provided by an embodiment of the present invention;

Fig. 4 is a structural block diagram of an equipment allocation system provided by an embodiment of the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

In the embodiment of the present invention, the system is powered on, and the physical basic input-output system (BasicInput-output System, BIOS) and graphics card BIOS are controlled to start, and the initialization operation is completed; the initialization setting of I/O access is controlled to the virtual system BIOS; through direct Storage access (Direct Memory Access, DMA) remapping hardware intercepts the transmission of the DMA of the device, and controls the address conversion, and the DMA operation of the control device directly accesses the memory space of the client; when the device with BIOS is allocated, by setting the client The machine shadow page table directly maps the I/O device to the MMIO space of the client, and the MMIO operation can directly access the hardware.

Figure 1 shows the implementation process of the device allocation method provided by the embodiment of the present invention, and its specific steps are as follows:

In step S101, the system is powered on, and the dual graphics cards are started in the physical BIOS to complete the initialization operation.

In step S102, the initialization setting of I/O access to the BIOS of the virtual system is controlled.

In the embodiment of the present invention, there are multiple ways to control the initial setting of I/O access to the virtual system BIOS, and the following embodiments provide three, but not to limit the present invention:

1. Hide the devices that need to be allocated from other clients other than the target client. The hidden method is to load a pseudo-driver on the allocated device before the driver program is installed on the client/host with the hardware device Placeholder, since there is no real driver, the device cannot be accessed by clients other than the intended client.

2. Control the virtual part monitor to intercept the virtual I/O address space accessed by the client, and forward the I/O request to the The real I/O address space of the device.

Pre-establish the conversion table between the virtual I/O address accessed by the client and the real I/O address of the device, and report the virtual Peripheral Component Interconnection (PCI) BAR to the client. When the client computer accesses the virtual I/O address space, the virtual machine monitor controls and intercepts the access operation, and forwards the I/O request to the real I/O address space of the device through the above translation table.

3. The virtual PCI configuration space generated for the device is attached to the virtual PCI bus.

Generate a virtual PCI configuration space for the device, attach the generated virtual PCI configuration space to the virtual PCI bus in the form of a virtual device, and when the guest operating system enumerates the bus, it can discover the device and load the correct driver.

In step S103, the DMA transmission of the device is intercepted by the DMA remapping hardware, and the address conversion is controlled to control the DMA operation of the device to directly access the memory space of the client computer.

In the embodiment of the present invention, the execution of the above steps is completed under the VT-d technology platform. As shown in FIG. 2 , the VT-d technology introduces DMA remapping hardware into the north bridge.

On the platform that starts VT-d, all DMA transfers of the device are intercepted by the DMA remapping hardware. According to the I/O page table corresponding to the device, the hardware can convert the DMA address so that the device can only access the specified memory .

In step S104, when the device with the BIOS is allocated, the I/O device is directly mapped to the MMIO space of the client by setting the shadow page table of the client, and the MMIO operation can directly access the hardware. In the embodiment of the present invention, due to the diversity of devices, it is impossible to provide a set of common BIOS for similar devices. Therefore, for devices with BIOS, the device BIOS area is directly mapped to the same area of the client address space, and then the client shadow page In the table, according to the correct page table entry written in the P2M table, the guest operating system can directly call the BIOS to control the hardware device.

As an embodiment of the present invention, as shown in Figure 3: all PCI configuration space access and data access operations of the guest operating system will be intercepted by the virtual machine monitor, forwarded to the device emulator, and then passed to the virtualized device The virtual machine monitor accesses hardware devices; for direct allocation devices, it directly accesses hardware data through MMIO (Memory Mapped I/O) and DMA; when a device with BIOS is directly allocated, the I/O device is directly allocated through the client shadow page table. Mapped to the MMIO space of the client, the MMIO operation of the client can directly access the hardware without forwarding by the virtual machine monitor; the DMA operation of the device can directly access the client address space through IOMMU (I/O Memory Management Unit) and reverse address translation Interaction; the BIOS of the device needs to be peer-to-peer mapped to the same area of the client address space, so that the guest operating system can call the device BIOS to access the device.

In order to solve the problem of direct allocation of hardware devices on the platform that does not support VT-d, an area can be set aside at the low end of the system memory. The guest machine page number is equal to the machine page number to solve the DMA problem of hardware access in the guest operating system, so that the guest operating system can directly access the physical device.

In the embodiment of the present invention, the virtual machine monitor cannot directly allocate those MMIO devices without page alignment, and for devices without page alignment, the virtual machine monitor is controlled to scan the PCI bus, and the address of the MMIO without page alignment is changed to Align the page.

In the virtual machine monitor, the memory used by the program code and the stack is located in the position of 1MB to 12MB of the physical memory. However, after the 1:1 per inch mapping, this section of memory may be accessed by the guest operating system. The operating system accesses this segment of memory, causing the system to crash, and moving the program code and stack of the virtual machine monitor to the 1:1 memory.

Fig. 4 shows the structural block diagram of the equipment distribution system provided by the embodiment of the present invention, for the convenience of explanation, only the part related to the embodiment of the present invention is shown in the figure, wherein, the equipment distribution system can be built in the software unit of the computer, A hardware unit or a combination of hardware and software.

Power on the system, the initialization operation module 11 starts the dual graphics card in the physical BIOS, and completes the initialization operation; the setting module 12 controls the initialization setting that the virtual system BIOS is carried out to I/O access; the DMA remapping control module 13 intercepts the device by the DMA remapping hardware The transmission of the DMA, and control address translation, the DMA operation of the control device directly accesses the memory space of the client computer; when the BIOS device allocation control module 14 allocates the device with the BIOS, by setting the client computer shadow page table, the I The /O device is directly mapped into the guest's MMIO space, which MMIO operations can directly access the hardware.

In the embodiment of the present invention, the modification module 15 controls the virtual machine monitor to scan the PCI bus for devices without page alignment, and controls to modify the MMIO addresses without page alignment to page alignment.

In the embodiment of the present invention, the hidden control module 121 hides the equipment that needs to be allocated from other clients except the target client; the address space conversion control module 122 controls the virtual component monitor to intercept the virtual I/O address space accessed by the client, According to the pre-established conversion table of the virtual I/O address accessed by the client and the real I/O address of the device, the I/O request is forwarded to the real I/O address space of the device; The PCI configuration space is attached to the virtual PCI bus.

The above is only a description of the system embodiment of the present invention, and its specific implementation is as described in the above process, which will not be repeated here, but it is not used to limit the present invention.

In the embodiment of the present invention, the system is powered on, the dual graphics cards are started in the physical BIOS, and the initialization operation is completed; the initialization setting of I/O access is controlled to the virtual system BIOS; the transmission of the DMA of the device is intercepted by the DMA remapping hardware, and Control the address conversion, and control the DMA operation of the device to directly access the memory space of the client; when the device with BIOS is allocated, the I/O device is directly mapped to the MMIO space of the client by setting the client shadow page table , the MMIO operations can directly access the hardware, realizing efficient virtualization and allocation of devices.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. within range.

Claims (8)

1. a device allocation method, is characterized in that, described method comprises the steps:

System powers on, and starts two video cards in physics BIOS, completes initialization operation;

The initialization setting of I/O access is carried out in control to virtual system BIOS;

Remap the transmission of the DMA of hardware capture device by DMA, and control and carry out address translation, mark a zone by the low side at Installed System Memory, it is distributed to client, client computer page number in client is equated with the machine page number, and the dma operation of opertaing device directly has access to the memory headroom of client computer;

Divide timing at the equipment with BIOS, by the client computer shadow page table is set, I/O equipment is mapped directly in the MMIO space of client computer, for the equipment that there is no page alignment, control monitor of virtual machine scanning pci bus, control and do not have the MMIO address of page alignment to change page alignment into by described, described MMIO operation is access hardware directly.

2. the method for claim 1, is characterized in that, the step that described control is carried out the initialization setting of I/O access to virtual system BIOS specifically comprises the steps:

Other client computer outside Destination client are hidden and needed the equipment distributed.

3. the method for claim 1, is characterized in that, the step that described control is carried out the initialization setting of I/O access to virtual system BIOS specifically comprises the steps:

Control virtual watch-dog and intercept and capture the virtual i/o address space of client access, according to the conversion table of the true I/O address of the virtual i/o address of setting up in advance client access and equipment, I/O is asked to the true I/O address space of the equipment that is forwarded to.

4. the method for claim 1, is characterized in that, the step that described control is carried out the initialization setting of I/O access to virtual system BIOS specifically comprises the steps:

To be articulated on virtual pci bus for the virtual pci configuration space that equipment generates.

5. a devices allocation system, is characterized in that, described devices allocation system is built in computing machine, and described system comprises:

The initialization operation module, power on for system, starts two video cards in physics BIOS, completes initialization operation;

Module is set, for controlling the initialization setting of virtual system BIOS being carried out to the I/O access;

DMA remaps control module, for remap the transmission of the DMA of hardware capture device by DMA, and control and carry out address translation, mark a zone by the low side at Installed System Memory, it is distributed to client, client computer page number in client is equated with the machine page number, and the dma operation of opertaing device directly has access to the memory headroom of client computer; And

BIOS devices allocation control module, for at the equipment with BIOS, dividing timing, by the client computer shadow page table is set, I/O equipment is mapped directly in the MMIO space of client computer, for the equipment that there is no page alignment, control monitor of virtual machine scanning pci bus, control and do not have the MMIO address of page alignment to change page alignment into by described, described MMIO operation is access hardware directly.

6. system as claimed in claim 5, is characterized in that, the described module that arranges specifically comprises:

Hide control module, for other client computer to outside Destination client, hide and need the equipment distributed.

7. system as claimed in claim 5, is characterized in that, the described module that arranges specifically comprises:

Address space conversion and control module, intercept and capture the virtual i/o address space of client access for controlling virtual watch-dog, according to the conversion table of the true I/O address of the virtual i/o address of setting up in advance client access and equipment, the I/O request is forwarded to the true I/O address space of equipment.

8. system as claimed in claim 5, is characterized in that, the described module that arranges specifically comprises:

The on-hook module, for being articulated in virtual pci bus for the virtual pci configuration space that equipment generates.

Images