WO2023165308A1 - Memory reclaim method and apparatus, and control device - Google Patents

Abstract

Provided in the present application are a memory reclaim method and apparatus, and a control device. The method comprises: acquiring memory management metadata information of a virtual machine, and a memory analytic function for analyzing the memory management metadata information; and on the basis of the memory analytic function and the memory management metadata information, reclaiming an idle memory of the virtual machine. By means of reclaiming the idle memory of a virtual machine, the problem that memory mapping of DMA on a device side cannot be changed is avoided, and an additional hardware or semi-virtualized driving program is not needed.

Description

A memory recovery method, device and control device

This application claims the priority of the Chinese patent application with the application number 202210197905.7 and the application name "A memory recovery method, device and control device" submitted to the China Patent Office on March 1, 2022, the entire content of which is incorporated by reference In this application.

technical field

The present application relates to the technical field of memory allocation, in particular, to a method, device and control device for memory recovery.

Background technique

Reclaiming the free memory of the virtual machine and improving the memory elasticity of the virtual machine can effectively improve the resource utilization efficiency of the host machine, help increase the density of virtual machines that the host machine can deploy, and reduce the cost of using virtual machines. However, in today's large-scale popularization of virtual machine device passthrough, device-side DMA requires that the memory map of the virtual machine cannot be changed at all during the running process, so it is impossible to dynamically reclaim the memory of the virtual machine.

One way to solve this problem is to provide device-side page fault handling capability, but this method depends on special hardware and cannot be used in existing hardware. In addition, there are some paravirtualization solutions that can solve such problems, but these solutions require a dedicated paravirtualization driver to run in the virtual machine.

Contents of the invention

The problem solved by this application is that the memory mapping of DMA cannot be changed in the existing memory recovery method, so that the memory of the virtual machine cannot be recovered.

In order to solve the above problems, this application firstly provides a method for memory recovery, including:

Obtaining memory management metadata information of the virtual machine and a memory parsing function for parsing the memory management metadata information;

Based on the memory analysis function and the memory management metadata information, reclaim the virtual machine free memory.

In one embodiment, after reclaiming the free memory of the virtual machine based on the memory analysis function and the memory management metadata information, the method further includes:

Allocating new physical memory to the virtual machine in response to the virtual machine's allocation request for the reclaimed free memory.

In an implementation manner, the memory parsing function is BPF bytecode compiled by the virtual machine.

In one embodiment, the memory analysis function for obtaining the memory management metadata information of the virtual machine and parsing the memory management metadata information includes:

When the pre-written eBPF source program of the memory analysis function is compiled after the virtual machine is started, the memory management metadata information injected by the virtual machine and the compiled BPF bytecode of the memory analysis function are received .

In one embodiment, based on the memory analysis function and the memory management metadata information, the free memory of the virtual machine is reclaimed, and scanned by the memory analysis function and the memory management metadata information After the free memory of the virtual machine is recovered, the free memory of the virtual machine is reclaimed.

In one embodiment, the memory management metadata information includes struct page physical page information and corresponding struct page structure information; After the free memory of the virtual machine is recovered, the free memory of the virtual machine is reclaimed, including:

Scanning the struct page physical page information of the virtual machine, calling the memory analysis function to analyze and correspond to the struct page structure information, and determining the idle struct page physical page;

Delete the free mapping of the struct page physical page in the EPT mapping table and the IOMMU mapping table, and make the mapping permission position of the struct page structure physical page in the EPT mapping table and the IOMMU mapping table read-only, and the struct page structure The body physical page is a struct page physical page that stores the struct page structure information corresponding to the free said struct page physical page.

In one embodiment, the allocation request of the virtual machine for the reclaimed free memory is a physical access to the struct page structure whose permission bit is read-only in the EPT mapping table or the IOMMU mapping table. Triggered when the page is mapped.

In one embodiment, the allocating new physical memory to the virtual machine is re-establishing the mapping of free struct page physical pages in the EPT mapping table or the IOMMU mapping table.

The second control device of the present application includes: a memory and a processor;

said memory for storing programs;

the processor, coupled to the memory, for executing the program for:

Obtaining memory management metadata information of the virtual machine and a memory parsing function for parsing the memory management metadata information;

Reclaim the free memory of the virtual machine based on the memory analysis function and the memory management metadata information.

In one embodiment, the processor is specifically configured to:

Allocating new physical memory to the virtual machine in response to the virtual machine's allocation request for the reclaimed free memory.

In an implementation manner, the memory parsing function is BPF bytecode compiled by the virtual machine.

In one embodiment, the processor is specifically configured to:

When the pre-written eBPF source program of the memory analysis function is compiled after the virtual machine is started, the memory management metadata information injected by the virtual machine and the compiled BPF bytecode of the memory analysis function are received .

In an implementation manner, the processor is specifically configured to reclaim the free memory of the virtual machine after scanning out the free memory of the virtual machine through the memory analysis function and the memory management metadata information.

In one embodiment, the memory management metadata information includes struct page physical page information and corresponding struct page structure information; based on this, the processor is specifically configured to:

Scanning the struct page physical page information of the virtual machine, calling the memory analysis function to analyze and correspond to the struct page structure information, and determining the idle struct page physical page;

Delete the mapping of the struct page physical page that is free in the EPT mapping table and the IOMMU mapping table, and map the mapping of the struct page structure physical page in the EPT mapping table and the IOMMU mapping table The permission location is read-only, and the physical page of the struct page structure is a physical page of the struct page that stores the information of the struct page structure corresponding to the free physical page of the struct page.

In one embodiment, the processor is specifically configured to:

The allocation request of the reclaimed free memory by the virtual machine is triggered when the virtual machine accesses the mapping of the physical page of the struct page structure whose permission bit is read-only in the EPT mapping table or the IOMMU mapping table.

In one embodiment, the processor is specifically configured to:

The allocating new physical memory to the virtual machine is to re-establish the mapping of free struct page physical pages in the EPT mapping table or the IOMMU mapping table.

The present application provides a memory recycling device again, which includes:

An information obtaining module, which is used to obtain memory management metadata information of the virtual machine and a memory analysis function for parsing the memory management metadata information;

A memory recycling module, configured to reclaim free memory of the virtual machine based on the memory analysis function and the memory management metadata information.

In one embodiment, the device also includes:

A memory allocation module, configured to allocate new physical memory to the virtual machine in response to the allocation request of the virtual machine for the reclaimed free memory.

This application reclaims the free memory of the virtual machine, thereby avoiding the problem that the device side DMA cannot change the memory mapping, and does not require additional hardware or paravirtualization drivers.

In this application, the virtual machine has no other changes except for injecting code and data into the host when it starts, so the intrusion control of the virtual machine is minimal, and it is easy to achieve software compatibility; the host actively and spontaneously updates the memory status of the virtual machine Scanning without waiting for the response of the virtual machine, so compared with the traditional PV solution, it can achieve better real-time performance; this solution is a pure software solution based on the existing architecture, and does not rely on supporting Page Fault on the device side Specific hardware and therefore hardware compatible.

In this application, memory management metadata information and memory management metadata analysis functions are injected into the host machine through the virtual machine, so that the host machine can scan and reclaim the memory of the virtual machine actively in real time, improving the response speed of memory reclamation; And the recycling/reallocation of free memory does not involve Disk I/O, Therefore, the processing efficiency is also improved; while recycling, the modification of the corresponding struct page mapping is considered, so that this solution can be used in device direct scenarios.

Description of drawings

FIG. 1 is a flowchart of a memory recovery method according to an embodiment of the present application;

FIG. 2 is a flowchart of a memory recovery method according to another embodiment of the present application;

FIG. 3 is a schematic diagram of a memory recovery process according to the present application;

FIG. 4 is a flow chart of a memory recovery method according to another embodiment of the present application;

Fig. 5 is the schematic diagram before reclaiming according to EPT mapping table and IOMMU mapping table according to the present application;

Fig. 6 is the schematic diagram after recycling according to the EPT mapping table and the IOMMU mapping table of the present application;

FIG. 7 is a structural block diagram of a memory recovery device according to an embodiment of the present application;

Fig. 8 is a structural block diagram of a control device according to an embodiment of the present application.

Detailed ways

In order to make the above purpose, features and advantages of the present application more obvious and understandable, specific embodiments of the present application will be described in detail below in conjunction with the accompanying drawings.

The existing memory reclamation method dynamically reclaims the memory of the virtual machine by changing the memory mapping, but the device-side DMA requires that the memory mapping of the virtual machine cannot be changed at all during the running process of the virtual machine.

Aiming at the above problems, the present application provides a solution to reclaim the free memory of the virtual machine by determining the free memory area of the virtual machine, so that there is no need to change the memory mapping being used during the running of the virtual machine.

For ease of understanding, the following terms that may be used are explained here:

Memory reclamation: After the host allocates memory to the virtual machine, it reclaims the excess memory when the virtual machine cannot make full use of it

Memory elasticity: the host can allocate the memory used by the virtual machine on demand

DMA: Direct Memory Access, which means that I/O devices can directly access main memory without CPU intervene

Device passthrough: In a virtualization scenario, a technology that allows the host hardware to be directly used by the VM

Device-side page fault interrupt: In a virtualization scenario, when the pass-through device accesses the memory of the virtual machine during DMA, if the host does not establish a mapping from the corresponding memory of the virtual machine to the memory of the host, the DMA will fail. In the case of hardware and software support, a page fault interrupt on the device side will be triggered, and the host will replay the DMA after the memory map is established.

Memory management metadata: metadata that records the status of each memory page frame in the operating system. In the Linux scenario, it is the memory area where the struct page is located and the content in the struct page

GVA: The virtual address space seen by the program inside the virtual machine

GPA: The physical memory space seen by the virtual machine

HPA: host physical memory space

EPT: Extended Page Table, used to establish a page table for mapping between virtual machine GPA and host machine HPA in a virtualization scenario

IOMMU: In a virtualization scenario, it is used to establish a page table for mapping between IOVA (generally virtual machine GPA) and HPA

Guest: Refers to the virtual machine

Host: Refers to the host

It should be noted that, under different operating environments, the representation form or corresponding structure of the memory will be different; in this application, the specific implementation mode under the Linux environment is taken as an example for illustration.

The embodiment of the present application provides a memory recovery method, which can be executed by a memory recovery device, and the memory recovery device can be integrated in electronic devices such as computers, servers, computers, server clusters, and data centers. As shown in FIG. 1, it is a flowchart of a memory recovery method according to an embodiment of the present application; wherein, the memory recovery method includes:

S100, acquiring memory management metadata information of the virtual machine and a memory analysis function for parsing the memory management metadata information;

S200. Reclaim free memory of the virtual machine based on the memory analysis function and the memory management metadata information.

In this embodiment, the virtual machine runs in the host machine, and one or more virtual machines can be set in one host machine, and different virtual machines are isolated from each other. The isolation between the same virtual machine is to prevent the resources (CPU, I/O device) allocated to other virtual machines from accessing the physical address of this virtual machine. Each virtual machine will have its own independent physical address space, that is, GPA (Guest Physical Address) space, which is different from the host physical address space, that is, HPA (Host Physical Address) space.

In one implementation manner, when there are multiple virtual machines, free memory of the virtual machine is reclaimed based on the memory analysis function corresponding to the virtual machine and the memory management metadata information.

Although the smallest addressable unit of the processor is usually a word (or even a byte), the linus kernel still regards the physical page as the basic unit of memory management, and uses the struct page structure to represent each physical page in the system; memory management is usually done in pages processed as a unit.

It should be noted that different architectures support different physical page sizes, and some architectures even support several different physical page sizes. Most 32-bit architectures support 4KB physical pages, while 64-bit architectures generally support 8KB physical pages.

Based on the size of the physical page and the size of the physical memory of the host machine, the physical memory can be divided into several physical pages. For example, on a host that supports a 4KB physical page size and has 1GB of physical memory, the physical memory will be divided into 262144 physical pages.

In the existing memory usage mode, in the GPA space of the virtual machine, physical pages with the same characteristics are generally gathered together, for example, free physical pages are placed together, so that when parsing free physical pages , the organization of the physical pages of the memory will give the number of physical pages. It is only necessary to analyze whether the head page (the first physical page) is a free physical page. Combined with the number of physical pages, all the free pages of the virtual machine can be determined. physical page.

It should be noted that for the GPA space of a virtual machine, its free physical pages are generally DMA buffers.

In this way, the free memory of the virtual machine is reclaimed, thereby avoiding the problem that the device side DMA cannot change the memory mapping, and no additional hardware or paravirtualization driver is required.

The embodiment of the present application provides another memory recovery method, which is similar to the aforementioned memory recovery method, except that, as shown in FIG. 2 , the S200 is based on the memory analysis function and The memory management metadata information, after reclaiming the free memory of the virtual machine, the method further includes:

S300. Allocate new physical memory to the virtual machine in response to the virtual machine's allocation request for the reclaimed free memory.

In this way, when the reclaimed free memory is re-allocated by the virtual machine for program running or DMA buffer, this part of the mapping is rebuilt by allocating the request, so it can effectively avoid the page fault on the device side caused by the DMA buffer not being mapped. .

In one embodiment, the free memory corresponding to the allocation request of the virtual machine is a part of all the free memory recovered from the virtual machine; in this case, when allocating new physical memory to the virtual machine, only The portion of memory requested by the virtual machine.

In one embodiment, when the corresponding physical memory is not occupied. Allocating new physical memory to the virtual machine is the physical memory reclaimed from the virtual machine. In this way, the original idle parameters of this part of the virtual machine can be restored to the greatest extent, thereby reducing the possible impact of memory reclamation on DMA and the like.

In an implementation manner, the memory parsing function is BPF bytecode compiled by the virtual machine. In this way, the compilation of the memory analysis function based on the eBPF program can be completed directly on the existing virtual machine, without the need for the virtual machine to provide additional hardware or additional paravirtualization drivers.

As shown in FIG. 3 , it is a schematic diagram of a memory recovery process, and the following content is described in detail in conjunction with this figure.

The embodiment of the present application provides another memory recovery method, which is similar to the above-mentioned memory recovery method, the difference is that in S100, the memory management metadata information of the virtual machine is obtained and the memory management metadata information is parsed memory parsing functions, including:

When the pre-written eBPF source program of the memory analysis function is compiled after the virtual machine is started, the memory management metadata information injected by the virtual machine and the compiled BPF bytecode of the memory analysis function are received .

In this embodiment, the eBPF source program of the memory analysis function is pre-written and put into the file system of the virtual machine; after the virtual machine starts, the eBPF source program of the memory analysis function in the file system is compiled into BPF bytecode, and The BPF bytecode of the compiled memory parsing function is injected into the host; through this In this way, the host machine obtains the BPF bytecode of the memory parsing function of the virtual machine.

In one embodiment, the memory management metadata information of the virtual machine includes struct page physical page information and corresponding struct page structure information; the struct page structure is created by the kernel for each struct page physical page, and the The struct page structure includes the field flags, which is used to describe the status and other information of the physical page. The field flags is mainly divided into 4 parts, in which the flag bit (the status identifier of the physical page) increases to the high bit, and the other bit fields section (mainly used Based on the sparse memory model SPARSEMEM), node (NUMA node number, which node the physical page belongs to), zone (memory domain mark, which zone the physical page belongs to) grows toward the lower bit, and there are idle bits in the middle.

In this embodiment, the memory management metadata information and memory analysis function will be injected into the host machine only after the virtual machine is started once, and the host machine can only scan the struct page physics of the virtual machine that has been injected with the memory management metadata information and memory analysis function Page. That is to say, the host machine can only reclaim the free memory of the virtual machine that has been started.

It should be noted that there can be multiple virtual machines corresponding to the host machine; at the same time, the host machine can only scan the struct page physical pages of the virtual machines that have been started. In addition, the virtual machine can only be started if the host has allocated memory.

For example, there are 11 virtual machines in the host machine, the physical memory of the host machine is 100G, and the memory to be allocated to the virtual machines is 10G. In this case, you can first select 10 virtual machines to allocate the corresponding 10G memory, and then start the The 10 virtual machines reclaim their free memory at the same time until the reclaimed free memory reaches 10G, and then distribute the reclaimed 10G memory to the 11th virtual machine, thereby completing the memory allocation work of all virtual machines.

The embodiment of the present application provides another memory recovery method, which is similar to the aforementioned memory recovery method, except that, as shown in Figure 4,

The S200, based on the memory analysis function and the memory management metadata information, reclaim the free memory of the virtual machine, and scan out the memory of the virtual machine through the memory analysis function and the memory management metadata information After free memory, reclaim the free memory of the virtual machine.

The memory management metadata information includes struct page physical page information and corresponding struct page structure information;

After the free memory of the virtual machine is scanned through the memory analysis function and the memory management metadata information, reclaiming the free memory of the virtual machine includes:

S201, scanning the struct page physical page information of the virtual machine, calling the memory parsing function to parse and correspond to the struct page structure information, and determining free struct page physical pages;

S202, delete the free mapping of the struct page physical page in the EPT mapping table and the IOMMU mapping table, and set the permission position of the mapping of the struct page structure physical page in the EPT mapping table and the IOMMU mapping table to read-only, and the struct The page structure physical page is a struct page physical page that stores the struct page structure information corresponding to the free struct page physical page.

In this way, the virtual machine only transmits its memory management metadata information and memory analysis function to the host at the initial stage of startup; after that, the host can use this information to actively scan the free memory area in the virtual machine and reclaim the free memory of the virtual machine; And while reclaiming the free memory of the virtual machine and clearing the memory map of the free page, the memory map of the memory management metadata is cleared, so as to ensure that the memory we reclaim must be the free memory in the virtual machine.

In one embodiment, call the memory analysis function to analyze and correspond to the struct page structure information, and determine the free struct page physical page, that is, identify the flag bit flag in the struct page structure by calling the memory analysis function, The struct page physical page corresponding to the flag bit flag of the preset identification is determined as a free struct page physical page.

In this embodiment, different virtual machines may have different labels for the flag bit flag in the struct page structure, so the flag bit flag for expressing "idle" is a preset flag and will also be different; for example, in a virtual machine, The flag bit in the struct page structure is PG-free, indicating that the corresponding struct page physical page is free; in another virtual machine, the flag bit flag in the struct page structure is PG-clean, indicating that the corresponding struct page is free page physical page.

In the present embodiment, delete the mapping of the free described struct page physical page in the EPT mapping table and the IOMMU mapping table, and the permission position of the mapping of the struct page structure physical page in the EPT mapping table and the IOMMU mapping table is read-only, The struct page structure physical page is a struct page physical page storing the struct page structure information corresponding to the idle struct page physical page.

As shown in FIG. 5 and FIG. 6 , the EPT mapping table and the IOMMU mapping table are relatively similar, so in this embodiment, the process of memory recovery is illustrated through the EPT mapping table. What needs to be emphasized here is that Figure 5 and Figure 6 show the EPT mapping table and IOMMU mapping table before and after recycling respectively. The schematic diagram after receipt, the figure only expresses the simple content of the general process, and does not strictly correspond to the specific content of the EPT mapping table and the IOMMU mapping table. Therefore, when understanding, some expressions that are not drawn in the figure should not be used To identify errors in drawings or representations.

As shown in the figure, in Figure 5, it can be seen that the mapping relationship between the struct page physical pages of GPA, HPA, and IOVA is recorded in the EPT mapping table and the IOMMU mapping table; and is used to describe the struct page The struct page structure of the physical page also needs to be stored, which is stored in another struct page physical page, and the struct page physical page storing the struct page structure is the above-mentioned struct page structure physical page. Among them, the struct page physical page showing the mapping relationship through the connection is considered as a free struct page physical page.

When reclaiming free struct page physical pages, delete the mapping of the free struct page physical pages in the EPT mapping table and IOMMU mapping table, that is, the struct page physical pages connected by the dotted line where Remove EPT Mapping and Remove IOMMU Mapping are located in Figure 6 Page. Set the permission location of the mapping of the physical page of the struct page structure in the EPT mapping table and the IOMMU mapping table to read-only, that is, set the struct page physical page of the storage location of the struct page structure corresponding to the struct page physical page that deletes the mapping The permission bit of the map is changed to "read-only".

In this embodiment, the permission bit is modified based on the struct page physical page as the basic unit, that is to say, once the permission bit of the mapping of a struct page physical page is modified to "read-only", the struct page All struct page structures stored in physical pages are restricted to "read-only".

In this embodiment, the struct page physical page has a certain capacity and can store multiple struct page structures. For example, a 4KB struct page physical page can store 64 struct page structures with a size of 64B; the 64 The struct page structure corresponds to 64 struct page physical pages. In this way, the modification of the permission bits of the physical page of the struct page structure in the EPT mapping table and the IOMMU mapping table means that the struct page structure is also modified to "read-only" at the same time, so the corresponding 64 The mapping of a struct page physical page in the EPT mapping table and the IOMMU mapping table.

It can be seen from this that the modification of the physical page of the struct page structure is linked with the struct page physical page corresponding to the physical page of the struct page structure, and needs to be kept in sync; therefore, if the 64 in the physical page of the struct page structure A struct page structure has a corresponding struct page physical page that is not free (even if the remaining 63 are free), then the struct page structure physical page and its The corresponding 64 struct page physical pages cannot be reclaimed; only the 64 struct page structure physical pages in the struct page structure physical pages are all idle struct page physical pages, and all of them can be reclaimed.

Similarly, if new physical memory needs to be allocated to the reclaimed virtual machine, even if the virtual machine only needs one struct page physical page, 64 struct page physical pages should be allocated at the same time.

In one embodiment, in the S300, the allocation request of the virtual machine to the reclaimed free memory is when the virtual machine accesses the struct page whose permission bit is read-only in the EPT mapping table or the IOMMU mapping table Triggered when the physical page of the structure is mapped.

Among them, if the virtual machine needs to use new memory, it will access the corresponding struct page structure in the physical page of the "read-only" struct page structure, but this access will exceed the "read-only" permission, and return A page-fault signal. When the host receives the page-fault signal, it can be considered that the virtual machine has given an allocation request for the reclaimed free memory.

In one embodiment, the allocating new physical memory to the virtual machine is re-establishing the mapping of free struct page physical pages in the EPT mapping table or the IOMMU mapping table.

In this way, the modification of the permission bit of the mapping of the physical page of the struct page structure is a solution determined after considering the execution path of the kernel memory allocation of the virtual machine. Through this modification, when the virtual machine allocates memory, its access to the struct page must be It will be intercepted by the host (returned page-fault signal), so that the page can establish EPT and IOMMU mapping before the virtual machine is available to prevent DMA failure.

In this application, the virtual machine has no other changes except for injecting code and data into the host when it starts, so the intrusion control of the virtual machine is minimal, and it is easy to achieve software compatibility; the host actively and spontaneously updates the memory status of the virtual machine Scanning without waiting for the response of the virtual machine, so compared with the traditional PV solution, it can achieve better real-time performance; this solution is a pure software solution based on the existing architecture, and does not rely on supporting Page Fault on the device side Specific hardware and therefore hardware compatible.

In this application, memory management metadata information and memory management metadata analysis functions are injected into the host machine through the virtual machine, so that the host machine can scan and reclaim the memory of the virtual machine actively in real time, improving the response speed of memory reclamation; Moreover, the recycling/redistribution of free memory does not involve Disk I/O, so the processing efficiency is also improved; while recycling, the modification of the corresponding struct page mapping is considered, so that this solution can be used in device direct scenarios.

An embodiment of the present application provides a memory reclamation device configured to execute the memory reclamation method described above in the present application. The memory reclamation device will be described in detail below.

As shown in Figure 7, the memory recycling device includes:

An information acquisition module 101, which is used to acquire memory management metadata information of a virtual machine and a memory analysis function for parsing the memory management metadata information;

A memory recycling module 102, configured to reclaim free memory of the virtual machine based on the memory analysis function and the memory management metadata information.

In one embodiment, the device also includes:

A memory allocation module 103, configured to allocate new physical memory to the virtual machine in response to the virtual machine's allocation request for the reclaimed free memory.

During the operation execution process of the information acquisition module 101, the memory analysis function is the compiled BPF bytecode of the virtual machine.

The information acquisition module 101 is also used for:

When the pre-written eBPF source program of the memory analysis function is compiled after the virtual machine is started, the memory management metadata information injected by the virtual machine and the compiled BPF bytecode of the memory analysis function are received .

The memory reclaiming module 102 is further configured to reclaim the free memory of the virtual machine after scanning out the free memory of the virtual machine through the memory analysis function and the memory management metadata information.

In one embodiment, the memory management metadata information includes struct page physical page information and corresponding struct page structure information; the memory recycling module 102 is also used for:

Scanning the struct page physical page information of the virtual machine, calling the memory analysis function to analyze and correspond to the struct page structure information, and determining the idle struct page physical page;

Delete the free mapping of the struct page physical page in the EPT mapping table and the IOMMU mapping table, and make the mapping permission position of the struct page structure physical page in the EPT mapping table and the IOMMU mapping table read-only, and the struct page structure The body physical page is a struct page physical page that stores the struct page structure information corresponding to the free said struct page physical page.

During the execution of the operation of the memory allocation module 103, the reclaimed idle memory of the virtual machine The memory allocation request is triggered when the virtual machine accesses the mapping of the physical page of the struct page structure whose permission bit is read-only in the EPT mapping table or the IOMMU mapping table.

During the operation execution process of the memory allocation module 103, the allocation of new physical memory to the virtual machine is to re-establish the mapping of free struct page physical pages in the EPT mapping table or the IOMMU mapping table.

The memory reclamation device provided by the above embodiments of the present application is based on the same inventive concept as the memory reclamation method provided by the embodiments of the present application, and has the same beneficial effect as the method adopted, run or implemented by the stored application program.

The above describes the internal function and structure of the memory recovery device, as shown in FIG. 8 , in practice, the memory recovery device can be implemented as a control device, including: a memory 301 and a processor 303 .

The memory 301 may be configured to store programs.

In addition, the memory 301 may also be configured to store other various data to support operations on the control device. Examples of such data include instructions for any application or method operating on the controlling device, contact data, phonebook data, messages, pictures, videos, etc.

The memory 301 can be implemented by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.

The processor 303, coupled to the memory 301, is used to execute the program in the memory 301 for:

Obtaining memory management metadata information of the virtual machine and a memory parsing function for parsing the memory management metadata information;

Reclaim the free memory of the virtual machine based on the memory analysis function and the memory management metadata information.

In one implementation manner, the processor 303 is specifically configured to:

Allocating new physical memory to the virtual machine in response to the virtual machine's allocation request for the reclaimed free memory.

In one embodiment, the memory parsing function is the compiled BPF byte of the virtual machine code.

In one implementation manner, the processor 303 is specifically configured to:

When the pre-written eBPF source program of the memory analysis function is compiled after the virtual machine is started, the memory management metadata information injected by the virtual machine and the compiled BPF bytecode of the memory analysis function are received .

In one implementation manner, the processor 303 is specifically configured to: reclaim the free memory of the virtual machine after scanning out the free memory of the virtual machine through the memory analysis function and the memory management metadata information.

In one embodiment, the memory management metadata information includes struct page physical page information and corresponding struct page structure information; based on this, the processor 303 is specifically used to:

Scanning the struct page physical page information of the virtual machine, calling the memory analysis function to analyze and correspond to the struct page structure information, and determining the free struct page physical page;

Delete the free mapping of the struct page physical page in the EPT mapping table and the IOMMU mapping table, and make the mapping permission position of the struct page structure physical page in the EPT mapping table and the IOMMU mapping table read-only, and the struct page structure The body physical page is a struct page physical page that stores the struct page structure information corresponding to the free said struct page physical page.

In one implementation manner, the processor 303 is specifically configured to:

The allocation request of the reclaimed free memory by the virtual machine is triggered when the virtual machine accesses the mapping of the physical page of the struct page structure whose permission bit is read-only in the EPT mapping table or the IOMMU mapping table.

In one implementation manner, the processor 303 is specifically configured to:

The allocating new physical memory to the virtual machine is to re-establish the mapping of free struct page physical pages in the EPT mapping table or the IOMMU mapping table.

In this application, only some components are schematically shown in FIG. 8 , which does not mean that the server device only includes the components shown in FIG. 8 .

The control device provided in this embodiment is based on the same inventive concept as the memory recovery method provided in the embodiment of this application, and has the same method adopted, run or implemented by its stored application program. Beneficial effect.

Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowcharts and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

Memory may include non-permanent storage in computer readable media, in the form of random access memory (RAM) and/or nonvolatile memory such as read-only memory (ROM) or flash RAM. Memory is an example of computer readable media.

Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can be implemented by any method or technology for storage of information. Information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory ( EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic tape cartridges, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media excludes transitory computer-readable media, such as modulated data signals and carrier waves.

It should also be noted that the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus comprising a set of elements includes not only those elements, but also includes Other elements not expressly listed, or elements inherent in the process, method, commodity, or apparatus are also included. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems or computer program products. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The above descriptions are only examples of the present application, and are not intended to limit the present application. For those skilled in the art, various modifications and changes may occur in this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims (11)

A memory recovery method, characterized in that, comprising: Obtaining memory management metadata information of the virtual machine and a memory parsing function for parsing the memory management metadata information; Reclaim the free memory of the virtual machine based on the memory analysis function and the memory management metadata information. The method according to claim 1, wherein after reclaiming the free memory of the virtual machine based on the memory analysis function and the memory management metadata information, the method further comprises: Allocating new physical memory to the virtual machine in response to the virtual machine's allocation request for the reclaimed free memory. The method according to claim 1, wherein the memory parsing function is BPF bytecode compiled by the virtual machine. The method according to claim 1, wherein the memory analysis function for obtaining the memory management metadata information of the virtual machine and parsing the memory management metadata information includes: When the pre-written eBPF source program of the memory analysis function is compiled after the virtual machine is started, the memory management metadata information injected by the virtual machine and the compiled BPF bytecode of the memory analysis function are received . The method according to claim 2, characterized in that, based on the memory analysis function and the memory management metadata information, the free memory of the virtual machine is reclaimed, and the memory analysis function and the memory After the management metadata information scans out the free memory of the virtual machine, reclaim the free memory of the virtual machine. The method according to claim 5, wherein the memory management metadata information includes struct page physical page information and corresponding struct page structure information; the memory analysis function and the memory management element After data information scans out the free memory of the virtual machine, reclaim the free memory of the virtual machine, including: Scanning the struct page physical page information of the virtual machine, calling the memory analysis function to analyze and correspond to the struct page structure information, and determining the idle struct page physical page; Delete the free mapping of the struct page physical page in the EPT mapping table and the IOMMU mapping table, and make the mapping permission position of the struct page structure physical page in the EPT mapping table and the IOMMU mapping table read-only, and the struct page structure The body physical page is a struct page physical page that stores the struct page structure information corresponding to the free said struct page physical page. The method according to claim 6, wherein the allocation request of the reclaimed free memory by the virtual machine is that the permission bit in the virtual machine accessing the EPT mapping table or the IOMMU mapping table is read-only Triggered when the physical page of the struct page structure is mapped. The method according to claim 6, wherein said allocating new physical memory to said virtual machine is re-establishing the mapping of free struct page physical pages in said EPT mapping table or said IOMMU mapping table. A control device, characterized in that it includes: a memory and a processor; said memory for storing programs; the processor, coupled to the memory, for executing the program for: Obtaining memory management metadata information of the virtual machine and a memory parsing function for parsing the memory management metadata information; Reclaim the free memory of the virtual machine based on the memory analysis function and the memory management metadata information. A memory recycling device is characterized in that it comprises: An information obtaining module, which is used to obtain memory management metadata information of the virtual machine and a memory analysis function for parsing the memory management metadata information; A memory recycling module, configured to reclaim free memory of the virtual machine based on the memory analysis function and the memory management metadata information. The device according to claim 10, further comprising: A memory allocation module, configured to allocate new physical memory to the virtual machine in response to the allocation request of the virtual machine for the reclaimed free memory.