WO2022227093A1

WO2022227093A1 - Virtualization system and method for maintaining memory consistency in virtualization system

Info

Publication number: WO2022227093A1
Application number: PCT/CN2021/091774
Authority: WO
Inventors: 胡雅琴; 李硕; 盖辰宁; 丁帅
Original assignee: 华为技术有限公司
Priority date: 2021-04-30
Filing date: 2021-04-30
Publication date: 2022-11-03
Also published as: CN116830093A

Abstract

Embodiments of the present application relate to the technical field of virtualization, and provide a virtualization system and a method for maintaining memory consistency in a virtualization system. The system comprises a request node and a first management node; the request node is used for sending a distributed virtual memory (DVM) request to the first management node when maintaining memory consistency of the virtualization system, the DVM request comprising broadcast range information; the first management node is used for parsing the DVM request to obtain the broadcast range information, the broadcast range information indicating information of M target nodes, and M being a positive integer; and the first management node is also used for sending broadcast information to each of the M target nodes, the broadcast information being used for instructing each target node to maintain the memory consistency. In this way, when the memory consistency is maintained, the broadcast range of an MN may be limited, the time that the MN waits for an invalid operation of a physical core is reduced, the physical core does not need to interrupt a task, and the performance of the physical core is not affected.

Description

Virtualized system and memory consistency maintenance method in virtualized system

technical field

The present application relates to the field of virtualization technology, and in particular, to a virtualization system and a method for maintaining memory consistency in the virtualization system.

Background technique

With the vigorous development of processors and related software ecosystems based on the advanced risc machine (ARM) architecture, the ARM architecture has been widely used in embedded, consumer electronics, big data or cloud computing and other fields. The ARM architecture supports a multi-core system, and a multi-core system can be understood as a processor of an electronic device that can support multiple physical cores.

Exemplarily, FIG. 1 shows a scenario diagram of an ARM architecture-based processor in a virtual machine (virtual machine, VM) application. As shown in Figure 1, this virtual machine application scenario can include four levels of domains, domain 0 includes one physical core, domain 1 includes four physical cores, domain 2 includes two domain 1s, and domain 3 includes Domain 2 and other domains (shown in the figure), wherein, domains 1 to 4 can all correspond to their own shared domains (shareable domains), and the shared domains of domains 1 to 4 can be used to store and schedule their respective running needs The specific content of the information depends on the actual application scenario, which is not limited here.

A virtual machine may correspond to multiple virtual central processing units (vCPUs). For example, in Figure 1, virtual machine #1 may correspond to vCPU1-vCPU4, and virtual machine #2 may correspond to vCPU5-vCPU8. The virtual machine kernel CPU scheduler can schedule one vCPU to run on multiple physical cores, or multiple vCPUs to run on one physical core. Figure 1 shows only one vCPU running on one physical core. An example of a physical core does not limit the physical core on which the vCPU runs. It can be understood that in the virtual machine application scenario of FIG. 1, in addition to virtual machine #1 and virtual machine #2, other virtual machines, such as VMx, etc., may also be included. The architecture of VMx is the same as that of virtual machine #1 or virtual machine #2. similar, and will not be repeated here.

As shown in Figure 1, the virtual machine application scenario also includes a storage (memory) system and other nodes (miscellaneous node, MN). The memory system is used to store memory page tables, etc. The memory page table can also be called a page table (page table), and the memory page table can store the mapping from virtual address (virtual address, VA) to physical address (physical address, PA), Among them, VA is unique in the access process, and PA is unique in hardware. In the virtual machine application scenario, each physical core first obtains the memory page table related to each physical core from the memory system, and each physical core stores the mapping of the obtained memory page table in its own translation lookaside buffer (translation lookaside buffer, TLB), when the physical core needs to obtain the mapping between VA and PA, the physical core first searches from its own TLB. If the physical core cannot search for the required mapping in its own TLB, the physical core can go to the memory system. search. When a physical core modifies the information of one or more mapping entries in the memory page table, the TLBs of other physical cores in the virtual machine application scenario need to synchronize the modified entries to maintain memory consistency .

The MN is used to manage multiple physical cores. For example, the MN can receive distributed virtual memory (DVM) requests from a certain physical core, and the DVM requests can be used to maintain memory consistency in virtual machine application scenarios. The MN may send a snoop (snp) DVM request to other physical cores, and the snpDVM request is used to instruct other physical cores to maintain the consistency of their respective TLBs. The MN receives response (Resp) information from other physical cores, and the MN returns Resp information to the physical core that initiated the DVM request. Among them, each physical core maintains TLB consistency can be understood as, each physical core invalidates a modified entry in the TLB, and subsequently, if the physical core needs to obtain the content of the invalid entry, the physical core obtains it from the memory system.

It should be noted that, the virtual machine application scenario of FIG. 1 can be applied to a certain electronic device, and the physical core, memory system and MN shown in FIG. 1 can all be included in the electronic device. The virtual machine application scenario shown in FIG. 1 can also be applied to a distributed system. The distributed system may include multiple electronic devices, and the physical core, memory system, and MN shown in FIG. 1 may be distributed and set in different electronic devices. . The MN may also be referred to as a management node, and the MN may be a physical core used to implement a management function, or multiple physical cores used to implement the management function. limited.

In a possible implementation, when each physical core maintains TLB consistency, an inter-processor interrupt (IPI) method may be used. Specifically, when a physical core modifies the entry information in the memory page table in the shared memory, in addition to the fact that the physical core needs to complete the invalidation operation of the entry information in its own TLB, the physical core also needs to send the MN to other physical cores. IPI instruction. The IPI instruction includes information indicating the maintenance of TLB consistency. Other physical cores invalidate their respective TLB entry information in the IPI interrupt processing function to achieve the purpose of maintaining shared memory page table consistency.

However, the IPI method is used in the above possible implementation. When the physical core receives the IPI request, it will suspend the currently executing task, respond to the interrupt, execute the page table synchronization and then return to the original program. This process will affect the physical core. performance.

SUMMARY OF THE INVENTION

The embodiments of the present application provide a virtualization system and a method for maintaining memory consistency in the virtualization system. When maintaining memory consistency, the broadcast range of the MN can be limited, the time for the MN to wait for an invalid operation of the physical core is reduced, and the physical core does not need to Interrupt tasks without affecting the performance of physical cores.

In a first aspect, an embodiment of the present application provides a virtualization system, including a request node and a first management node; the request node is configured to send a distributed virtual system to the first management node when performing memory consistency maintenance of the virtualization system. Memory DVM request, the DVM request includes broadcast range information; the first management node is used to parse the DVM request to obtain the broadcast range information, and the broadcast range information indicates the information of M target nodes; M is a positive integer; It is used to send broadcast information to each of the M target nodes; the broadcast information is used to instruct each target node to perform memory consistency maintenance.

Based on this, when maintaining memory consistency in the virtualization system, the embodiments of the present application can limit the broadcast range of the first management node, reduce the time that the first management node waits for the invalid operation of the physical core, and the physical core does not need to interrupt the task, Does not affect the performance of physical cores.

In a possible implementation manner, the broadcast range information includes: an identifier of a physical core related to a virtual machine running in the requesting node, and/or an identifier of a layer related to the virtual machine running in the requesting node; wherein the layers correspond to multiple physical core. In this way, when the broadcast range information in this embodiment of the present application indicates the broadcast range, the physical core can be used as the granularity to indicate the broadcast range to achieve precise broadcast range limitation, and the broadcast range can also be indicated by the layer as the granularity. The identifier of one layer can correspond to multiple Therefore, the indication of multiple physical cores can be realized through fewer identifiers, which saves system resources.

In a possible implementation manner, the first management node is specifically configured to send broadcast information to each of the M target nodes when the identifier of the M physical cores indicated by the broadcast range information is parsed; or, The first management node is specifically configured to send broadcast information to each of the M target nodes in the layer when the identifier of the layer indicated by the broadcast range information is parsed. In this way, the broadcast range of the first management node can be limited, and the time for the first management node to wait for the invalid operation of the physical core can be reduced.

In a possible implementation manner, the nodes indicated by the broadcast range information include nodes managed by the second management node; the first management node is further configured to send a DVM request to the second management node; the second management node is configured to The DVM request is parsed to obtain broadcast range information, and broadcast information is sent to nodes managed by the second management node. In this way, when the second management node sends broadcast information to a node managed by the second management node, accurate broadcast range limitation can be achieved.

In a possible implementation manner, the layers indicated by the broadcast range information include a layer managed by the second management node; the first management node is further configured to send a DVM request to the second management node; the second management node is configured to The DVM request is parsed to obtain the broadcast range information, and the broadcast information is sent to the nodes in the layer managed by the second management node. In this way, when the second management node sends the broadcast information to the nodes in the layer managed by the second management node, the indication to multiple nodes can be implemented by using fewer identifiers, thereby saving system resources.

In a possible implementation manner, the requesting node is provided with a register; the register is used to store broadcast range information; the requesting node is specifically used to generate an instruction when performing memory consistency maintenance of the virtualization system, and combine the instruction with the broadcast range The information is packaged as a DVM request, and the DVM request is sent to the first management node. In this way, by setting a register in the requesting node to store the broadcast range information, the broadcast range can be updated by definition, and the requesting node can package the instruction and the broadcast range information into a DVM request to send to the first management node, which is convenient and quick.

In one possible implementation, the instruction includes a translation lookaside buffer instruction TLBI or a cache maintenance instruction IC instruction. In this way, the translation lookaside buffer instruction TLBI can maintain the consistency of the TLB in the physical core, and the cache maintenance instruction IC instruction can also maintain the consistency of the TLB in the physical core.

In a possible implementation manner, the first management node is further configured to collect M response information from M target nodes; the response information of each target node is used to indicate that the target node completes memory consistency maintenance; the first management node , and is also used to send information indicating the completion of memory consistency maintenance to the requesting node. In this way, the first management node can collect and transmit information indicating that the memory consistency maintenance is completed.

In a possible implementation manner, the requesting node is further configured to generate a data synchronization isolation DSB instruction when multiple DVM requests are sent within a preset time, and the DSB instruction is used to instruct the first management node to collect and complete multiple DVM requests After receiving the response information from the corresponding node, the information for indicating the completion of the memory consistency maintenance is synchronously sent to the first management node. In this way, the requesting node can generate a DSB instruction, indicating that the maintenance of the memory consistency of the nodes corresponding to the multiple DVM requests is completed.

In a possible implementation manner, the first management node is specifically configured to send broadcast information to each of the M target nodes in a covering manner; wherein the covering manner is to cover nodes other than the M target nodes The way. In this way, the physical cores of other nodes except the target node can be covered, and the other nodes will not receive the broadcast information, so the waiting time of the first management node can be saved, and the calculation amount of the system can be reduced.

In a second aspect, an embodiment of the present application provides a method for maintaining memory consistency in a virtualization system, including: when a requesting node performs memory consistency maintenance in a virtualization system, sending a distributed virtual memory DVM request to a first management node, The DVM request includes broadcast range information; the first management node parses the DVM request to obtain broadcast range information, and the broadcast range information indicates the information of M target nodes; M is a positive integer; the first management node sends a message to each of the M target nodes. Send broadcast information; broadcast information is used to instruct each target node to maintain memory consistency.

In a possible implementation manner, the broadcast range information includes: an identifier of a physical core related to a virtual machine running in the requesting node, and/or an identifier of a layer related to the virtual machine running in the requesting node; wherein the layers correspond to multiple physical core.

In a possible implementation manner, the first management node sends broadcast information to each of the M target nodes, including: when the first management node parses the identifiers of the M physical cores indicated by the broadcast range information, Send broadcast information to each of the M target nodes; or, when the first management node parses the identifier of the layer indicated by the broadcast range information, sends broadcast information to each of the M target nodes in the layer. .

In a possible implementation manner, the nodes indicated by the broadcast range information include nodes managed by the second management node, and/or layers managed by the second management node, and the method further includes: the first management node sends the second management node to the second management node. The management node sends a DVM request.

In a possible implementation manner, the requesting node sends a distributed virtual memory DVM request to the first management node when performing the memory consistency maintenance of the virtualization system, including: the requesting node is performing the memory consistency maintenance of the virtualization system When , an instruction is generated; the requesting node packages the instruction and the broadcast range information into a DVM request; the requesting node sends the DVM request to the first management node.

In one possible implementation, the instruction includes a translation lookaside buffer instruction TLBI or a cache maintenance instruction IC instruction.

In a possible implementation manner, the first management node collects M pieces of response information from M target nodes; the response information of each target node is used to indicate that the target node has completed memory consistency maintenance; the first management node sends a message to the requesting node Information used to indicate that memory consistency maintenance is complete.

In a possible implementation manner, when the requesting node sends multiple DVM requests within a preset time, a data synchronization isolation DSB instruction is generated, and the DSB instruction is used to instruct the first management node to collect and complete the data of the nodes corresponding to the multiple DVM requests. After responding to the information, synchronously sends to the first management node information indicating that the memory consistency maintenance is completed.

In a possible implementation manner, the first management node sends broadcast information to each of the M target nodes, including: the first management node sends broadcast information to each of the M target nodes in a covering manner; The covering method is a method of covering nodes other than the M target nodes.

In a third aspect, an embodiment of the present application provides a method for maintaining memory consistency in a virtualization system, including: a first management node receives a distributed virtual memory DVM request from a requesting node, where the DVM request includes broadcast range information; the first management node Parse the DVM request to obtain broadcast range information, the broadcast range information indicates the information of M target nodes; M is a positive integer; the first management node sends broadcast information to each of the M target nodes; the broadcast information is used to indicate each The target node performs memory consistency maintenance.

In a fourth aspect, an embodiment of the present application provides a method for maintaining memory consistency in a virtualized system, including: a requesting node generates an instruction when performing memory consistency maintenance in a virtualization system; the requesting node sets the instruction sum in the requesting node The broadcast range information is packaged into a DVM request; the requesting node sends the DVM request to the first management node.

In a possible implementation manner, the requesting node receives information from the first management node indicating that the maintenance of memory consistency is completed.

In a fifth aspect, embodiments of the present application provide a computer-readable storage medium, where the computer-readable storage medium includes a computer program, and when the computer program is run on an electronic device, the electronic device is made to perform any of the above-mentioned third and fourth aspects. A possible design of the technical solution.

A sixth aspect is a computer program product according to an embodiment of the present application. The computer program product includes instructions, and when the instructions are executed on a computer, the computer can execute the technical solutions of any of the third and fourth aspects.

For the beneficial effects of the second to sixth aspects, please refer to the beneficial effects of the first aspect, which will not be repeated.

Description of drawings

1 is a schematic diagram of a scenario of an ARM architecture-based processor in a virtual machine application provided by an embodiment of the present application;

2 is a schematic diagram of a first system architecture to which the method of the embodiment of the present application is applied;

3 is a schematic diagram of a second system architecture to which the method of the embodiment of the present application is applied;

FIG. 4 is a schematic diagram of a third system architecture to which the method of the embodiment of the present application is applied;

FIG. 5 is a schematic diagram of a fourth system architecture to which the method of the embodiment of the present application is applied;

6 is a schematic flowchart of data synchronization in a multi-core system provided by an embodiment of the present application;

7 is a schematic diagram of a logical architecture of an MN node parsing a DVM request according to an embodiment of the present application;

FIG. 8 is a schematic flowchart of a specific data synchronization method in a multi-core system according to an embodiment of the present application.

Detailed ways

In order to clearly describe the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish the same or similar items with basically the same function and effect. For example, the first event and the second event are only for distinguishing different events, and do not limit their order. Those skilled in the art can understand that the words "first", "second" and the like do not limit the quantity and execution order, and the words "first", "second" and the like are not necessarily different.

It should be noted that, in this application, words such as "exemplary" or "for example" are used to represent examples, illustrations or illustrations. Any embodiment or design described in this application as "exemplary" or "such as" should not be construed as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "such as" is intended to present the related concepts in a specific manner.

In this application, "at least one" means one or more, and "plurality" means two or more. "And/or", which describes the association relationship of the associated objects, indicates that there can be three kinds of relationships, for example, A and/or B, which can indicate: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A, B can be singular or plural. The character "/" generally indicates that the associated objects are an "or" relationship. "At least one item(s) below" or similar expressions thereof refer to any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (a) of a, b, or c can represent: a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, c may be single or multiple .

To facilitate understanding of the embodiments of the present application, some words involved in the embodiments of the present application are briefly described below.

1. Physical core: It can be manufactured from single crystal silicon with a certain production process, and is used to perform steps such as calculation, receiving or storing commands, and processing data. Each physical core has its own independent TLB.

2. TLB maintenance instruction (TLBI): an instruction used to instruct to maintain the consistency of the TLB in the physical core.

3. Virtual machine: refers to a complete computer system with complete hardware system functions simulated by software and running in an isolated environment.

4. Bus: Refers to the public communication trunk for transmitting data between various nodes, which can be used to transmit messages or requests.

5. Broadcast range register: refers to the register used to store broadcast range information.

6. Request node (RN): A physical core can correspond to an RN, or it can be understood that an RN is a node that runs a vCPU in the physical core to achieve a certain function. It should be noted that RN and MN are relative concepts, one MN can manage multiple RNs, and the multiple RNs can correspond to one virtual machine or multiple virtual machines.

The methods of the embodiments of the present application can be applied to virtual machine application scenarios in the fields of embedded, consumer electronics, big data, automotive electronics, mass storage, imaging equipment, industrial control, security systems, or cloud computing. In the virtual machine application scenario, the electronic device that executes the method of the embodiment of the present application includes a processor, and the processor may adopt an ARM architecture, and the processor based on the ARM architecture has the advantages of high speed, low power consumption, and low price.

The electronic device may also be referred to as a terminal device, a terminal (terminal), a user equipment (UE), a mobile station (mobile station, MS), or a mobile terminal (mobile terminal, MT). The electronic device can be a mobile phone (mobile phone), a smart TV, a wearable device, a tablet computer (Pad), a computer with a wireless transceiver function, a virtual reality (virtual reality, VR) terminal device, an augmented reality (augmented reality, AR) terminal Equipment, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in remote medical surgery, wireless terminals in smart grid, transportation Wireless terminals in security (transportation safety), wireless terminals in smart cities, wireless terminals in smart homes, and so on. The embodiments of the present application do not limit the specific technology and specific device form adopted by the electronic device.

With reference to FIG. 1 , in another possible implementation of maintaining memory consistency, the way of broadcasting invalid operation of TLB can be adopted. Specifically, when a physical core modifies one or more mapping entries in the memory page table in the shared memory, the physical core generates a TLBI. While completing the TLB invalidation operation of the physical core, the physical core also Package the TLBI as a DVM request, and send the DVM request to the MN through the coherent hub interface (CHI). After the MN receives the DVM request, the ARM software will open a force broadcast (FB) control instruction by default. , this instruction will force the MN to broadcast the DVM request to all physical cores within the inner shareable (IS) range, and the MN broadcasts the snpDVM request to all the physical cores within the IS range. The request invalidates the corresponding entry information in the respective TLB. After the MN collects the Resp information of all physical cores within the IS range, it indicates the completion of the TLB maintenance to the physical core that sent the DVM request, so as to maintain the TLB consistency of each physical core. Purpose. It should be noted that, in a common implementation, the IS range is defined at the beginning of the system design, and the IS range is large and cannot be changed.

Among them, in order to ensure that the snpDVM request takes effect synchronously on all physical cores within the IS scope, the MN can perform a DVM synchronization (synchronization, Sync) operation, so that the corresponding entry information in the completion TLB is invalid for all physical cores within the IS scope. After that, the Resp information is returned to the MN at the same time.

However, in the above implementation, the MN needs to indiscriminately wait for the corresponding entry information invalidation operation in the TLBs of all physical cores within the IS range to end, and there is a long waiting time and a large performance overhead.

In view of this, embodiments of the present application provide a virtualization system and a method for maintaining memory consistency in the virtualization system. When maintaining memory consistency, the broadcast range of the MN can be limited, the time for the MN to wait for an invalid operation of a physical core can be reduced, and The physical core does not need to interrupt tasks and does not affect the performance of the physical core. Specifically, when the RN in the embodiment of the present application sends a request to the MN, the request includes broadcast range information, that is, the broadcast range of the MN is reduced. After the MN sends the broadcast information to the node corresponding to the broadcast range information, it waits for each The time adaptation of node synchronization is shortened, the efficiency of maintaining consistency is improved, and each node does not need to interrupt its own tasks, which does not affect the performance of each node.

Wherein, a broadcast range register may be set in the physical core of the embodiment of the present application, and broadcast range information is preset in the broadcast range register. The broadcast range register can receive user settings. For example, the user can set or modify the broadcast range information in the broadcast range register by using CPU scheduler software in advance.

In this embodiment of the present application, the MN is configured with logic capable of parsing the broadcast range information requested by the DVM. In a possible implementation, a class Affinity register may be set in the MN, and the Affinity register may be set to indicate the RN corresponding to the MN and the system level (ie, layer) where the RN is located. The Affinity register is statically configurable, for example, it can be configured when software deploys a VM.

It should be noted that, when the broadcast range information in this embodiment of the present application indicates the broadcast range, the broadcast range may be indicated with a physical core as the granularity. For example, the broadcast range information may include the identifier of the physical core that needs to be broadcast. Broadcast to the physical core indicated by the broadcast range information. In a possible implementation, a broadcast range register with a range of 128 bits (binary digit, Bit) can be defined, and each Bit represents a physical core. When indicating the broadcast range of the MN according to the broadcast range information in the broadcast range register, the physical The core indicates the broadcast range for granularity, and the MN can broadcast the physical core according to the indication of the broadcast range information. It can be understood that, in this implementation, because the broadcast range information is based on the physical core as the granularity, accurate broadcast range limitation can be achieved.

When the broadcast range information in this embodiment of the present application indicates the broadcast range, the broadcast range may also be indicated by a layer as a granularity. One layer may include multiple physical cores, and the layer division rules are not limited in this embodiment of the present application. For example, the broadcast range information can include the identifier of the layer that needs to be broadcast, and when the MN broadcasts, it broadcasts all physical cores in the layer indicated by the broadcast range information. It can be understood that, in this implementation, because the broadcast range information is granular in layers, the identifier of one layer can correspond to multiple physical cores, so the indication of multiple physical cores can be implemented with fewer identifiers, saving system resources.

It can be understood that, in the embodiment of the present application, when the nodes indicated in the broadcast range information are all managed by one MN, the number of MNs in the embodiment of the present application may be one. When the node indicated in the broadcast range information is managed by multiple different MNs, the RN can send the broadcast range information to one of the MNs. In addition to sending the broadcast information to the node managed by the MN, the MN can also send the broadcast range information to other MNs. MN, so that the other MN can also send broadcast information to the node managed by the other MN.

Exemplarily, FIG. 2 to FIG. 5 show schematic diagrams of four possible virtualization system architectures according to the embodiments of the present application.

FIG. 2 shows a schematic diagram of a first virtualization system architecture to which the method of the embodiment of the present application is applied. In the schematic diagram of the virtualization system framework in FIG. 2 , when the broadcast range information indicates the broadcast range, the broadcast range is indicated with the physical core as the granularity, and the nodes indicated in the broadcast range information are all managed by one MN.

As shown in FIG. 2, the virtualization system includes RN-0, MN, and RN-1 to RN-n.

Wherein, RN-0 corresponds to physical core #1, TLB and broadcast range register are set in physical core #1, and broadcast range information is preset in the broadcast range register. Exemplarily, the broadcast range information may be stored in the broadcast range register through the virtual machine core CPU scheduler in advance, where the broadcast range information may include the identifier of the physical core related to the VM running in the RN-0. It can be understood that, when the broadcast range information in the broadcast register needs to be updated, the update code can be scheduled by the virtual machine core CPU scheduler to update the broadcast register.

The process of maintaining memory consistency in a virtualization system is described below by taking RN-0 modifying the entry information of the shared page table and initiating maintaining the consistency of entry information in the TLB of each physical core as an example.

After RN-0 modifies the entry information of the shared page table, TLBI is generated in RN-0, and the TLBI is used to maintain the TLB in physical core #1. RN-0 completes the invalidation operation of entry information in its own TLB, and further Yes, RN-0 can also obtain the updated memory page table from the memory system, set it in the TLB of RN-0, and complete the update of the shared page table of RN-0.

In addition, RN-0 can package the broadcast range information and TLBI into a DVM request, then the DVM request includes the broadcast range information, RN-0 sends the DVM request to the MN through the CHI bus, and the MN parses the broadcast range information to obtain the broadcast range The physical core indicated by the information, and the physical core indicated by the broadcast range information may also be referred to as other RNs.

Exemplarily, as shown in FIG. 2 , the physical core indicated by the broadcast range information may be RN-1, and the MN broadcasts a snpDVM request to RN-1. It can be understood that the number of physical cores indicated by the broadcast range information may be determined according to an actual scenario, and may be one or more, which is not specifically limited in this embodiment of the present application.

RN-1 can invalidate the entry information in its own TLB based on the snpDVM request, and RN-1 can also further obtain the updated memory page table from the memory system, and send Resp information to the MN.

In a possible implementation, RN-0 may modify multiple entry information in the memory page table within a period of time, then RN-0 may generate multiple TLBIs asynchronously within this period of time. If applicable, RN -0 Asynchronously sends the DVM requests corresponding to the multiple TLBIs to the MN. In order to ensure that the multiple entry information invalid operations indicated by multiple DVM requests are synchronously effective in the physical cores corresponding to the broadcast range indication information, RN-0 can generate data synchronization The isolation (data synchronization barrier, DSB) instruction instructs the MN to collect the Resp information of the nodes corresponding to the multiple DVM requests, and then synchronously sends the Resp information of all nodes to the RN-0, indicating the memory consistency of the nodes corresponding to the multiple DVM requests Maintenance is complete.

Exemplarily, FIG. 3 shows a schematic diagram of a second virtualization system architecture to which the method of this embodiment of the present application is applied. In the schematic diagram of the virtualization system framework in FIG. 3 , when the broadcast range information indicates the broadcast range, the physical core is used as the granularity to indicate the broadcast range, and the nodes indicated in the broadcast range information are managed by multiple different MNs.

As shown in FIG. 3, the virtualization system includes RN-0, MN1, MN2, RN1-1 to RN1-n, and RN2-1 to RN2-n. Among them, MN1 is used to manage RN1-1 to RN1-n, and MN2 is used to manage RN2-1 to RN2-n.

Wherein, RN-0 corresponds to physical core #1, TLB and broadcast range register are set in physical core #1, and broadcast range information is preset in the broadcast range register.

From the process of RN-0 modifying the entry information of the shared page table to the process of RN-0 sending DVM request to MN1 node, please refer to the description of RN-0 sending DVM request to MN node in the embodiment corresponding to FIG. Repeat.

Different from the embodiment corresponding to FIG. 2 , in the embodiment corresponding to FIG. 3 , the nodes indicated in the broadcast range information are managed by multiple different MNs. For example, the physical cores indicated in the broadcast range information include RN1-1, RN1 -2, RN2-1 and RN2-2, RN1-1 and RN1-2 are managed by MN1, and RN2-1 and RN2-2 are managed by MN2.

As shown in Figure 3, RN-0 sends a DVM request to MN1, MN1 parses the broadcast range information, and determines that the broadcast range information includes not only RN1-1 and RN1-2 corresponding to MN1, but also RN2-1 corresponding to MN2 and RN2-2.

MN1 broadcasts the snpDVM request to RN1-1 and RN1-2. At the same time, MN1 sends the DVM request to MN2. MN2 parses the broadcast range information and determines that the physical cores indicated in the broadcast range information include RN2-1 and RN2-2. RN2-1 and RN2-2 send snpDVM requests.

RN1-1 and RN1-2 may perform the entry information invalidation operation in the TLB and send Resp information to MN1 in the manner described in the embodiment corresponding to FIG. 2 .

RN2-1 and RN2-2 may perform the entry information invalidation operation in the TLB and send Resp information to MN2 in the manner described in the embodiment corresponding to FIG. 2 .

After MN2 collects the Resp information of RN2-1 and RN2-2, it can send the Resp information to MN1, and MN1 collects the Resp information of RN1-1, RN1-2 and MN2, and sends the Resp information to RN-0.

It can be understood that FIG. 2 shows the situation that the physical core indicated by the broadcast range information corresponds to two MNs. In practical applications, the number of MNs may be greater than 2, and each MN can transmit DVM requests to each other. The RN managed by each MN indicated in the range information broadcasts and collects reply information. The embodiment of the present application does not limit the number of MNs and the communication mode between multiple MNs.

Exemplarily, FIG. 4 shows a schematic diagram of a third virtualization system architecture to which the method of this embodiment of the present application is applied. In the schematic diagram of the virtualization system framework in FIG. 4 , when the broadcast range information indicates the broadcast range, the broadcast range is indicated by layers as granularity, and the nodes in the layers indicated in the broadcast range information are all managed by one MN.

As shown in FIG. 4, the virtualization system includes RN-0, MN0, RN0-1 to RN0-n. Among them, RN0-1 to RN0-n belong to one layer 0, and MN0 is used to manage RN0-1 to RN0-n.

Different from the embodiment corresponding to FIG. 2 , in the embodiment corresponding to FIG. 4 , when the broadcast range information indicates the broadcast range, the broadcast range is indicated by the granularity of layers. For example, layer 0 is indicated in the broadcast range information, and layer 0 includes RNO-1 to RNO-n.

As shown in FIG. 4 , RN-0 sends a DVM request to MN0, MN0 parses the broadcast range information, determines that the broadcast range information is layer 0, and MN0 sends snpDVM requests to RN0-1 to RN0-n in layer 0.

RNO-1 to RNO-n may perform the entry information invalidation operation in the TLB in the manner described in the corresponding embodiment of FIG. 2, and send Resp information to MNO, and MNO collects the Resp of RNO-1 to RNO-n in layer 0 information, and send Resp information to RN-0.

Exemplarily, FIG. 5 shows a schematic diagram of a fourth virtualization system architecture to which the method of this embodiment of the present application is applied. In the schematic diagram of the virtualization system framework in FIG. 5 , when the broadcast range information indicates the broadcast range, the broadcast range is indicated by layers as granularity, and multiple layers indicated in the broadcast range information are managed by multiple different MNs.

As shown in FIG. 5, the virtualization system includes RN-0, MN3, MN4, RN3-1 to RN3-n, and RN4-1 to RN4-n. RN3-1 to RN3-n belong to layer 3, RN4-1 to RN4-n belong to layer 4, MN3 is used to manage RN3-1 to RN3-n, and MN4 is used to manage RN4-1 to RN4-n.

Different from the embodiment corresponding to FIG. 4 , in the embodiment corresponding to FIG. 5 , the number of layers indicated in the broadcast range information is multiple, and the multiple layers are managed by different MNs. For example, the layers indicated in the broadcast range information include

layers

3 and 4, the nodes of layer 3 are managed by MN3, and the nodes of layer 4 are managed by MN4.

As shown in Figure 5, RN-0 sends the DVM request to MN3, MN3 parses the broadcast range information, and determines that the broadcast range information includes not only RN3-1 to RN3-n in layer 3 corresponding to MN3, but also the corresponding information of MN4. RN4-1 to RN4-n within layer 4.

MN3 broadcasts the snpDVM request to RN3-1 to RN3-n, and MN3 sends the DVM request to MN4, MN3 parses the broadcast range information, determines that the broadcast range information is RN3-1 to RN3-n in layer 3, and MN3 sends the layer In 3, RN3-1 to RN3-n broadcast the snpDVM request.

RN3-1 to RN3-n may perform the entry information invalidation operation in the TLB and send Resp information to MN3 in the manner described in the embodiment corresponding to FIG. 2 .

RN4-1 to RN4-n may perform the entry information invalidation operation in the TLB and send Resp information to MN4 in the manner described in the embodiment corresponding to FIG. 2 .

After MN4 collects the Resp information of RN4-1 to RN4-n, it can send the Resp information to MN3, and MN3 collects the Resp information of RN3-1 to RN3-n and MN4, and sends the Resp information to RN-0.

It can be understood that FIG. 5 shows the situation where there are two MNs corresponding to the layer indicated by the broadcast range information. In practical applications, the number of MNs may be greater than 2. Each MN can transmit DVM requests to each other to realize the broadcast range. Nodes in layers corresponding to each MN indicated in the information broadcast and collect reply information. The embodiment of this application does not limit the number of MNs and the communication mode between multiple MNs.

It should be noted that, taking Figure 2 as an example, the broadcast range of the MN itself may cover a large number of nodes such as RN-1 to RN-n, but due to the limitation of the broadcast range information, the MN can broadcast the snpDVM request to RN-1 , instead of sending snpDVM requests to RN-2 to RN-n, when maintaining page table consistency, MN does not need to wait for the replies from RN-2 to RN-n, which can save waiting time, improve maintenance efficiency, and reduce The amount of calculation of the system can save computing resources.

It can be understood that, in the embodiments corresponding to FIGS. 3-5 , similar to the description in FIG. 2 , the waiting time of the MN can be reduced, the maintenance efficiency can be improved, the calculation amount of the system can be reduced, and computing resources can be saved.

It should be noted that, in the embodiments corresponding to FIG. 2 to FIG. 5, TLBI can also be replaced with instructions such as cache maintenance instruction (instruction cache maintenance instruction, IC), and the IC instruction is used to implement the above functions of TLBI, which will not be repeated here. . In Figure 2-Figure 5, the reason for triggering memory consistency maintenance can also be that RN-0 modifies the entry information in its own TLB. For example, if RN-0 clears the entry information in its own TLB, then RN-0 0 can also request the relevant physical cores in the virtualization system to perform memory consistency maintenance. The process of memory consistency maintenance is detailed in the above description, and will not be repeated here.

In the embodiment corresponding to FIG. 2 to FIG. 5 , when the MN broadcasts the snpDVM instruction to the node indicated by the broadcast range information, it can be sent in a covering manner, wherein the covering method is to cover the physical cores of the nodes other than those indicated by the broadcast range information, The physical cores of nodes other than those indicated by the broadcast range information cannot receive the snpDVM instruction.

The communication method provided by the embodiments of the present application will be described in detail below with reference to the accompanying drawings. FIG. 6 shows a schematic flowchart of data synchronization in a multi-core system provided by the present application. As shown in FIG. 6 , the method of the embodiment of the present application includes:

S601. Request the node to generate an instruction.

In this embodiment of the present application, the instruction may be used to instruct the first management node to perform a certain operation, which may be the smallest functional unit of operation. For example, the instruction can be TLBI, or it can be an instruction such as IC.

Exemplarily, taking the requesting node as RN-0 running on physical core #1 as an example, when RN-0 modifies the entry information in the shared page table, or RN-0 modifies the entry information in its own TLB , RN-0 can generate instructions that can be used to instruct the maintenance of memory coherency of the virtualized system.

S602. The requesting node packages the instruction and broadcast range information into a DVM request.

In this embodiment of the present application, the broadcast range information is used to indicate the broadcast range of the first management node. For example, the broadcast range information may be used to indicate the nodes that the first management node needs to broadcast, and so on.

The broadcast range information may be set in a register, or may be set in other storage devices, which is not specifically limited in this embodiment of the present application.

In this embodiment of the present application, the DVM request is used to request the first management node to maintain the consistency of the TLB in the physical core. The DVM request carries broadcast range information.

Exemplarily, the RN-0 running on the physical core #1 can package the broadcast range information and the TLBI into a DVM request according to the TLBI.

S603. The requesting node sends a DVM request to the first management node.

Suitably, the first management node receives the DVM request from the requesting node.

In the embodiment of the present application, the possible implementation of the requesting node sending the DVM request to the first management node is as follows: the requesting node sends the DVM request to the first management node through the bus, if applicable, the first management node can receive the DVM request from the requesting node through the bus. ask.

It can be understood that, in the embodiment of the present application, it is recognized in the bus protocol that the DVM request includes the broadcast range information, so the bus can transmit the DVM request including the broadcast range information.

Exemplarily, taking the requesting node as the RN-0 running on the physical core #1 and the first management node as the MN, when performing page table consistency maintenance, the RN-0 sends a DVM request to the MN through the CHI bus, and the MN A DVM request from RN-0 is received through the CHI bus, and the DVM request includes broadcast range information.

S604. The first management node parses the broadcast range information to obtain M target nodes, where M is a positive integer.

In this embodiment of the present application, the target node refers to the node where the physical core indicated by the broadcast range information is located.

Exemplarily, taking the first management node as the MN as an example, after the MN receives the DVM request, the MN can parse the broadcast range information, and can obtain M nodes that need to be broadcast indicated by the broadcast range information, as M target nodes. .

S605. The first management node sends broadcast information to the M target nodes.

In this embodiment of the present application, the broadcast information is used to instruct the target node to perform memory consistency maintenance, for example, the broadcast information may be the snpDVM request in the above embodiment.

The memory consistency maintenance may refer to keeping the data in the target node consistent with the data in the first management node. For example, after RN-0 modifies the shared page table, the target node RN-1, the target node RN-2, and the target node RN-3 need to be updated to the modified shared page table in their respective TLBs. After receiving the broadcast information from the first management node, the M target nodes can perform memory consistency maintenance respectively. For example, any target node can invalidate the entry information currently stored in the TLB, access the updated shared page table from the memory system again, and store the updated shared page table in the TLB.

S606. The first management node collects M pieces of response information from the M target nodes.

In this embodiment of the present application, the response information is used to indicate that the target node completes data synchronization.

Exemplarily, after the target node RN-1 completes data synchronization, RN-1 sends response information to the MN, and adaptively, the MN collects the response information from RN-1. Similarly, the first management node may collect the response information from the target node RN-2 until the MN has collected the response information from the M target nodes.

It can be understood that the time at which each target node performs data synchronization may be different. Therefore, the timing at which each target node sends response information to the first management node may also be different. During this process, the first management node needs to wait until the M is collected. The response information of each node, the first management node can confirm that the memory consistency maintenance is completed.

In a possible implementation manner, the first management node may also forward the DVM request to the second management node. For example, in the scenario corresponding to FIG. 3 or FIG. 5, the second management node may be the MN2 node or the MN4 node, then the second management node may receive the DVM request from the first management node, and collect the corresponding data of the second management node according to the DVM request. the target node. For details, reference may be made to the explanations in FIG. 3 and FIG. 5 , which will not be repeated here.

S607: The first management node sends information for indicating completion of data synchronization to the requesting node.

Suitably, the requesting node receives information from the first management node indicating that the maintenance of memory consistency is completed.

In this embodiment of the present application, the information used to indicate the completion of memory consistency maintenance may be response information sent by the first management node to the requesting node to indicate the completion of memory consistency maintenance after collecting response information from the M target nodes. The information used to indicate that the memory consistency maintenance is completed may be in the form of numbers or characters, which is not specifically limited in this embodiment of the present application.

The requesting node can receive the information indicating that the memory consistency maintenance is completed, and confirm that the memory consistency maintenance is completed.

To sum up, in this embodiment of the present application, when the requesting node sends a request to the first management node, the request includes the broadcast range information, which can narrow the broadcast range of the first management node, and the subsequent request node sends a request to the corresponding broadcast range information. Sending broadcast information by a node can reduce the time that the requesting node waits for invalid operations of the node corresponding to the broadcast range information, improve the efficiency of maintaining memory consistency, and the node corresponding to the broadcast range information does not need to interrupt the task, and does not affect the node corresponding to the broadcast range information. performance.

On the basis of the embodiment corresponding to FIG. 6 , in a possible implementation manner, the broadcast range information may include: an identifier of a physical core related to a virtual machine running in the requesting node, and/or a virtual machine running in the requesting node The identification of the relevant layer; wherein, the layer corresponds to multiple physical cores.

In this embodiment of the present application, the number of virtual machines running in the requesting node may be one or multiple.

The number of VM-related physical cores may be one or multiple. The identifier of the physical core is used to clearly identify the physical core. For example, the identifier of the physical core may be the serial number, address, or name of the physical core, which is not specifically limited in this embodiment of the present application.

The number of VM-related layers may be one or multiple. The identifier of the layer is used to clearly identify the layer. For example, the identifier of the layer may be the serial number or name of the layer, which is not specifically limited in this embodiment of the present application.

The broadcast range information may include the identifier of the physical core related to the virtual machine running in the request node, or the broadcast range information may include the identifier of the layer related to the virtual machine running in the request node, or the broadcast range information may include the request The identification of the physical core and the identification of the layer related to the virtual machine running in the node. Then the subsequent first management node may parse the broadcast range information, and send the broadcast information to the physical core or layer indicated by the broadcast range information.

Optionally, the broadcast range information may be set in a register of the requesting node. The register may be a specially designed broadcast range register for storing broadcast range information, or may be any register in the requesting node, which is not specifically limited in this embodiment of the present application.

Exemplarily, FIG. 7 shows a schematic diagram of a logical architecture of an MN parsing a DVM request. As shown in FIG. 7, a register 701 can be set in the MN700.

After the MN receives the DVM request carrying the broadcast range information from the requesting node (sco physical core), it can compare the broadcast range information delivered by the bus with the node managed by the MN indicated in the register 701. If the broadcast range information corresponds to the MN If the node matches, the broadcast information will be forwarded to the matching node. The broadcast information may be, for example, a snpDVM request, and the snpDVM request includes a DVM code corresponding to the broadcast range information.

In another possible implementation, bitmaps (bitmaps) of physical cores may be set in the MN, and each bitmap indicates a corresponding RN.

After receiving the DVM request carrying the broadcast range information from the requesting node, the MN forwards the broadcast information to the RN specified by the bitmap information according to the bitmap information, for example, the broadcast information may be the snpDVM request.

On the basis of the embodiment corresponding to FIG. 6 , in a possible implementation manner, S605 includes: the first management node sends broadcast information to M target nodes in a covering manner, wherein the covering manner is The way the node is covered.

The covering mode may refer to covering the unused physical cores when the first management node sends the broadcast request.

In the embodiment of the present application, taking the first management node as the MN, the nodes corresponding to the MN include node 1-node m, and the target node includes node 1, node 5, and node 6 as an example, the MN may divide the node 1-node m, except The physical cores of other nodes except node 1, node 5 and node 6 are blocked, and broadcast information is sent. Because the physical cores of the other nodes are blocked, the other nodes will not receive the broadcast information, and the MN will not receive the broadcast information. It is necessary to wait for the reply of the other node about maintaining the page table consistency, so the waiting time of the MN can be saved, and the calculation amount of the system can be reduced.

Exemplarily, FIG. 8 is a schematic flowchart of a specific data synchronization method in a multi-core system provided by the present application. As shown in FIG. 8 , the method of the embodiment of the present application includes:

S801, the software maintains the broadcast range register.

In this embodiment of the present application, software refers to a series of sets of data and instructions organized in a specific order. For example, the software may be CPU scheduler software.

Exemplarily, the user can schedule the physical core #N running the VM through the CPU scheduler software, and pass the relevant physical core (that is, the broadcast range information) of the running of the VM through the custom instruction set architecture (instruction set architecture, ISA) instruction set. The broadcast range register of physical core #N is updated, so that the broadcast range register includes the broadcast range information.

S802. The requesting node generates a DVM request, and sends the DVM request through the bus.

Exemplarily, the requesting node modifies the shared page table, or it is understood as the VM software update translation, the requesting node may generate a TLBI, the requesting node packages the TLBI and the broadcast range information into a DVM request, and sends the DVM request to the MN based on the bus payload. The broadcast range information may also be understood as a custom broadcast range domain.

S803, the MN directionally forwards the broadcast.

Exemplarily, after receiving the DVM request, the MN parses the broadcast range information in the DVM request, and the MN sends the broadcast information to the RN indicated by the broadcast range information in a targeted manner by covering.

S804, the MN collects Resp information in a targeted manner.

Exemplarily, the MN directionally collects the Resp information of the RN indicated by the broadcast range information, and sends the Resp information to the requesting node after collecting the Resp information of the RN indicated by the broadcast range information.

In the embodiment of the present application, the CPU scheduler software maintains the broadcast range register, and can update the broadcast range by self-definition. Subsequently, the MN sends broadcast information to the corresponding node by covering it, which can reduce the MN waiting for the invalid operation of the corresponding node. Time, reduce software maintenance costs and the difficulty of MN broadcasting, and improve the efficiency of maintenance consistency.

It can be understood that the data synchronization method of the embodiment of the present application can also be easily transplanted to other physical cores of the ARM architecture for implementation, which not only does not increase the design difficulty of the MN, but also transfers the maintenance scope function to the software, which can reduce the cost of the MN. Hardware burden.

Embodiments of the present application also provide a computer-readable storage medium. The methods described in the above embodiments may be implemented in whole or in part by software, hardware, firmware or any combination thereof. If implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media can include both computer storage media and communication media and also include any medium that can transfer a computer program from one place to another. The storage medium can be any target medium that can be accessed by a computer.

As one possible design, the computer readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium intended to carry or in an instruction or data structure The required program code is stored in the form and can be accessed by the computer. Also, any connection is properly termed a computer-readable medium. For example, if you use coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies (such as infrared, radio, and microwave) to transmit software from a website, server, or other remote source, coaxial cable, fiber optic cable , twisted pair, DSL or wireless technologies such as infrared, radio and microwave are included in the definition of medium. Disk and disc as used herein includes compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and blu-ray disc, where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

The embodiments of the present application also provide a computer program product. The methods described in the above embodiments may be implemented in whole or in part by software, hardware, firmware or any combination thereof. If implemented in software, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the above-mentioned computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the above-mentioned method embodiments are generated. The aforementioned computer may be a general purpose computer, a special purpose computer, a computer network, a base station, a terminal, or other programmable devices.

The above specific embodiments further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific embodiments of the present invention, and are not intended to limit the protection scope of the present invention. On the basis of the technical solutions of the present invention, any modifications, equivalent replacements, improvements, etc. made shall be included within the protection scope of the present invention.

Claims

A virtualization system, characterized in that the virtualization system includes a request node and a first management node;

the requesting node, configured to send a distributed virtual memory DVM request to the first management node when performing memory consistency maintenance of the virtualization system, where the DVM request includes broadcast range information;

the first management node, configured to parse the DVM request to obtain the broadcast range information, where the broadcast range information indicates information of M target nodes; the M is a positive integer;

The first management node is further configured to send broadcast information to each of the M target nodes; the broadcast information is used to instruct each target node to perform memory consistency maintenance.
The virtualization system according to claim 1, wherein the broadcast range information includes: an identifier of a physical core related to a virtual machine running in the requesting node, and/or a virtual machine running in the requesting node machine-related layer identifiers; wherein, the layers correspond to multiple physical cores.
The virtualization system according to claim 2, wherein,

The first management node is specifically configured to send the broadcast information to each of the M target nodes when the identifiers of the M physical cores indicated by the broadcast range information are parsed;

Or, the first management node is specifically configured to send the broadcast information to each of the M target nodes in the layer when the identifier of the layer indicated by the broadcast range information is parsed.
The virtualization system according to any one of claims 2 or 3, wherein the nodes indicated by the broadcast range information include nodes managed by the second management node;

the first management node, further configured to send the DVM request to the second management node;

The second management node is configured to parse the DVM request to obtain the broadcast range information, and send the broadcast information to the node managed by the second management node.
The virtualization system according to any one of claims 2 or 3, wherein the layer indicated by the broadcast range information includes a layer managed by the second management node;

the first management node, further configured to send the DVM request to the second management node;

The second management node is configured to parse the DVM request to obtain the broadcast range information, and send the broadcast information to the nodes in the layer managed by the second management node.
The virtualization system according to any one of claims 1-5, wherein the request node is provided with a register; the register is used to store the broadcast range information;

The requesting node is specifically configured to generate an instruction when performing memory consistency maintenance of the virtualization system, package the instruction and the broadcast range information into the DVM request, and send the request to the first management node. Send the DVM request.
The virtualization system according to claim 6, wherein the instruction comprises a translation lookaside buffer instruction TLBI or a cache maintenance instruction IC instruction.
The virtualization system according to any one of claims 1-7, wherein,

The first management node is further configured to collect M response information from the M target nodes; the response information of each target node is used to indicate that the target node has completed memory consistency maintenance;

The first management node is further configured to send, to the requesting node, information indicating completion of memory consistency maintenance.
The virtualization system according to claim 8, wherein the requesting node is further configured to generate a data synchronization isolation DSB instruction when multiple DVM requests are issued within a preset time, and the DSB instruction is used to instruct the After collecting and completing the response information of the nodes corresponding to the multiple DVM requests, the first management node synchronously sends the information for indicating that the memory consistency maintenance is completed to the first management node.
The virtualization system according to any one of claims 1-9, wherein,

The first management node is specifically configured to send broadcast information to each of the M target nodes in a covering manner; wherein the covering method is to cover nodes other than the M target nodes. Way.
A method for maintaining memory consistency in a virtualization system, characterized in that the method includes:

When performing memory consistency maintenance of the virtualization system, the requesting node sends a distributed virtual memory DVM request to the first management node, where the DVM request includes broadcast range information;

The first management node parses the DVM request to obtain the broadcast range information, where the broadcast range information indicates information of M target nodes; the M is a positive integer;

The first management node sends broadcast information to each of the M target nodes; the broadcast information is used to instruct each target node to perform memory consistency maintenance.
The method according to claim 11, wherein the broadcast range information comprises: an identifier of a physical core related to a virtual machine running in the requesting node, and/or a related virtual machine running in the requesting node The identifier of the layer; wherein, the layer corresponds to multiple physical cores.
The method according to claim 12, wherein the first management node sends broadcast information to each of the M target nodes, comprising:

The first management node sends the broadcast information to each of the M target nodes when it parses the identifiers of the M physical cores indicated by the broadcast range information;

Alternatively, the first management node sends the broadcast information to each of the M target nodes in the layer when it parses that the broadcast range information indicates an identifier of a layer.
The method according to claim 12 or 13, wherein the nodes indicated by the broadcast range information include a node managed by a second management node, and/or a layer managed by the second management node, so The method also includes:

The first management node sends the DVM request to the second management node.
The method according to any one of claims 11-14, wherein the requesting node sends a distributed virtual memory DVM request to the first management node when performing memory consistency maintenance of the virtualization system, comprising: :

The requesting node generates an instruction when performing the memory consistency maintenance of the virtualization system;

the requesting node packages the instruction and the broadcast range information into the DVM request;

The requesting node sends the DVM request to the first management node.
16. The method of claim 15, wherein the instruction comprises a translation lookaside buffer instruction TLBI or a cache maintenance instruction IC instruction.
The method according to any one of claims 11-16, further comprising:

The first management node collects M response information from the M target nodes; the response information of each target node is used to indicate that the target node completes memory consistency maintenance;

The first management node sends information for indicating completion of memory consistency maintenance to the requesting node.
The method of claim 17, further comprising:

When the requesting node sends out multiple DVM requests within a preset time, a data synchronization isolation DSB instruction is generated, and the DSB instruction is used to instruct the first management node to collect and complete the responses of the nodes corresponding to the multiple DVM requests. After the information is received, synchronously sends the information indicating that the memory consistency maintenance is completed to the first management node.
The method according to any one of claims 11-18, wherein the first management node sends broadcast information to each of the M target nodes, comprising:

The first management node sends broadcast information to each of the M target nodes in a covering mode, wherein the covering mode is a mode of covering nodes other than the M target nodes.
A method for maintaining memory consistency in a virtualization system, characterized in that the method includes:

The first management node receives a distributed virtual memory DVM request from the requesting node, where the DVM request includes broadcast range information;

The first management node parses the DVM request to obtain the broadcast range information, where the broadcast range information indicates information of M target nodes; the M is a positive integer;

The first management node sends broadcast information to each of the M target nodes; the broadcast information is used to instruct each target node to perform memory consistency maintenance.
The method according to claim 20, wherein the broadcast range information comprises: an identifier of a physical core related to a virtual machine running in the requesting node, and/or a related virtual machine running in the requesting node The identifier of the layer; wherein, the layer corresponds to multiple physical cores.
The method according to claim 21, wherein the first management node sends broadcast information to each of the M target nodes, comprising:

The first management node sends the broadcast information to each of the M target nodes when it parses the identifiers of the M physical cores indicated by the broadcast range information;

Alternatively, the first management node sends the broadcast information to each of the M target nodes in the layer when it parses that the broadcast range information indicates an identifier of a layer.
The method according to claim 21 or 22, wherein the nodes indicated by the broadcast range information include a node managed by a second management node, and/or a layer managed by the second management node, so The method also includes:

The first management node sends the DVM request to the second management node.
The method according to any one of claims 20-23, further comprising:

The first management node collects M response information from the M target nodes; the response information of each target node is used to indicate that the target node completes memory consistency maintenance;

The first management node sends information for indicating completion of memory consistency maintenance to the requesting node.
The method according to any one of claims 20-24, wherein the first management node sends broadcast information to each of the M target nodes, comprising:

The first management node sends broadcast information to each of the M target nodes in a covering mode, wherein the covering mode is a mode of covering nodes other than the M target nodes.
A method for maintaining memory consistency in a virtualization system, characterized in that the method includes:

The requesting node generates an instruction when performing the memory consistency maintenance of the virtualization system;

The requesting node packages the instruction and the broadcast range information set in the requesting node into the DVM request;

The requesting node sends the DVM request to the first management node.
The method according to claim 26, wherein the broadcast range information comprises: an identifier of a physical core related to a virtual machine running in the requesting node, and/or a related virtual machine running in the requesting node The identifier of the layer; wherein, the layer corresponds to multiple physical cores.
The method of claim 26 or 27, wherein the instruction comprises a translation lookaside buffer instruction TLBI or a cache maintenance instruction IC instruction.
The method according to any one of claims 26-28, wherein,

The requesting node receives information from the first management node indicating that the maintenance of memory consistency is completed.
The method of claim 29, further comprising:

When the requesting node sends out multiple DVM requests within a preset time, a data synchronization isolation DSB instruction is generated, and the DSB instruction is used to instruct the first management node to collect and complete the responses of the nodes corresponding to the multiple DVM requests. After the information is received, synchronously sends the information indicating that the memory consistency maintenance is completed to the first management node.