Apparatus for transferring mediation requests for processing circuits in response to configuration requests - Patents.com
A mediation interface in data processing systems resolves conflicts by authorizing configuration requests from lower-privileged software through higher-privileged software, enhancing efficiency and versatility by eliminating direct software communication and reducing development costs.
Patent Information
- Application Number
- JP2021552859
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-03-12
- Filing Date
- 2020-01-03
- Publication Date
- 2024-12-05
- Estimated Expiration
- 2040-01-03
AI Technical Summary
In data processing systems with virtualized hardware devices, conflicts arise due to multiple software applications sharing resources without awareness of each other's usage, leading to inefficiencies and potential interference, especially when lower-privileged applications send configuration requests that may negatively impact other applications.
An interface in the processing circuitry mediates configuration requests from lower-privileged software by forwarding them to higher-privileged software for authorization, eliminating the need for direct communication and specific protocols between software applications, ensuring compatibility and versatility across different operating systems and platforms.
This approach reduces processing time and development costs by enabling transparent mediation of configuration requests, improving system efficiency and versatility without the need for specific drivers or protocol agreements between software applications, while maintaining system security and reducing the likelihood of conflicts.
Smart Images

Figure 0007595020000001 
Figure 0007595020000002 
Figure 0007595020000003
Abstract
Description
[Technical field]
[0001] The present technology relates to data processing systems. Summary of the Invention [Problem to be solved by the invention]
[0002] Technical background In a data processing system, dedicated hardware devices may be provided to perform specific tasks, and in some situations there may be multiple users or multiple software applications sharing access to the same hardware device. The device may be virtualized so that each user or software application believes that it is the only one using the device, but in reality the resources within the device are partitioned to be shared among multiple users.
[0003] Viewed from one aspect, the present technology provides an apparatus, comprising: a processing circuit configured to execute the software; and an interface configured to receive a configuration request from the processing circuitry from the first software requesting configuration of the virtualization device; In response to the configuration request, the interface is configured to forward the mediation request to the processing circuit; The intermediation request includes a request that a second software having a higher privilege level than the first software determine a response to a configuration request received from the first software.
[0004] Viewed from another aspect, the present technology provides a method, the method comprising: executing software on a processing circuit; receiving a configuration request from the processing circuitry from the first software requesting configuration of a virtualization device; and in response to the configuration request, forwarding the mediation request to a processing circuit; The intermediation request includes a request that a second software having a higher privilege level than the first software determine a response to a configuration request received from the first software.
[0005] Viewed from another aspect, the present technology provides an apparatus comprising: a means for executing the software; means for receiving a configuration request from the means for executing software, the configuration requesting configuration of a virtualization device from the first software; In response to the configuration request, forwarding the mediation request to the means for executing the software; The intermediation request includes a request that a second software having a higher privilege level than the first software determine a response to a configuration request received from the first software.
[0006] Viewed from another aspect, the present technology provides a computer program for controlling a host data processing apparatus to provide an instruction execution environment, the computer program comprising: processing program logic configured to execute the software; and interface program logic configured to receive a configuration request from the first software executed on the processing program logic, the configuration request requesting configuration of the virtualization device; In response to the configuration request, the interface program logic is configured to forward the mediation request to a second software executed in the processing program logic; the second software has a higher privilege level than the first software; The mediation request includes a request for the second software to determine a response to a configuration request received from the first software.
[0007] Further aspects, features and advantages of the present technology will become apparent from the following description of examples which should be read in conjunction with the accompanying drawings. [Brief description of the drawings]
[0008] [Figure 1] FIG. 1 illustrates an example hierarchy of privilege levels in a data processing system. [Diagram 2] FIG. 2 shows an example approach to mediation of configuration requests. [Diagram 3] FIG. 3 illustrates an alternative approach to mediation of configuration requests, in accordance with an embodiment of the present technology. [Figure 4] FIG. 4 shows an example of the configuration of a processing circuit and hardware devices. [Diagram 5] FIG. 5 illustrates an example of a data processing system. [Figure 6] FIG. 6 is a flow diagram illustrating an example of a method of operation of the management interface. [Figure 7A] FIG. 7A is a timing diagram illustrating an example of signals transferred between portions of a data processing system. [Figure 7B] FIG. 7B is a timing diagram illustrating an example of signals transferred between portions of a data processing system. [Figure 8] FIG. 8 shows an example of a management interface. [Figure 9] FIG. 9 shows an example of a register used in the management interface. [Figure 10] FIG. 10 is a flow diagram illustrating an example of a method of operation of the management interface. [Figure 11] FIG. 11 is a flow diagram illustrating an example of a method of operation of the processing circuit. [Figure 12A] FIG. 12A is a flow diagram illustrating the method involved in setting an intermediation enabled or disabled state. [Figure 12B] FIG. 12B is a flow diagram illustrating the method involved in setting an intermediation enabled or disabled state. [Figure 13] FIG. 13 shows an example of the use of a data processing system. [Figure 14] FIG. 14 shows an example of the use of a data processing system. [Figure 15] FIG. 15 shows an example of the use of a data processing system. [Figure 16] FIG. 16 shows an example of a simulator implementation. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0009] In a virtualization system, the virtualization device may receive instructions from multiple software applications executed by processing circuitry, the processing circuitry being circuitry configured to perform processing operations corresponding to the processing instructions, including processing instructions representing the software applications. The multiple software applications may all execute on the same processing circuitry, or may instead execute on different sets of processing circuitry (e.g., the processing circuitry may include multiple processor cores in a multi-core system). Each software application includes a privilege level that determines the particular set of permissions associated with that application, and the level of knowledge that application has about other software that accesses the virtualization device. In this way, a software application that runs at a lower privilege level, when issuing instructions to the virtualization device, may be unaware of other applications that use the resources of the virtualization device. This increases the likelihood of conflicts occurring between applications that request use of the virtualization device.
[0010] To avoid collisions, for some types of configuration commands for configuring a virtualization device, a higher privilege application may mediate requests sent to the virtualization device by lower privilege applications. For example, if a lower privilege application wants to send a configuration request to a virtualization device, requesting configuration of the virtualization device according to the requirements of the lower privilege application, the higher privilege program may mediate this request to ensure that it is not allowed if it could adversely affect other applications using the virtualization device. Such configuration requests are triggered in some examples by processing circuitry executing dedicated instructions to trigger the configuration request. However, in other examples, the configuration requests are triggered by processing circuitry executing load or store instructions that specify a memory address mapped to the configuration request. For example, writing to a particular memory address may trigger hardware to issue a configuration request to the virtualization device. Such requests and responses may be performed in an application-specific manner using writes to certain addresses or data structures in memory based on a pre-agreed protocol for interaction between low privilege software and high privilege software.
[0011] In one way of performing such mediation, the first (lower privilege) software may send a request to the second (higher privilege) software to mediate a configuration request that the first software wishes to send to the virtualization device, and the second software may then return a response to the first software indicating whether the first software has permission to send the configuration request, and if permission is granted, the first software may issue the configuration request.
[0012] This approach has the advantage of allowing for mediation of configuration requests, but the disadvantage of having a performance impact as well as requiring specific drivers to allow each software application to communicate with other software applications (using application-specific protocols for interaction between the applications known at the time the software was written). Thus, when a software application is ported to a different physical platform running a different operating system or other higher privileged code, this increases the development cost for the software developer of the ported application, since a different protocol is required for requesting mediation from the higher privileged code. This approach also incurs a cost for both the developer of the higher privileged code and the developer of the lower privileged code in that both must interact and agree on a protocol by which requests for mediation are sent from the lower privileged code to the higher privileged code and corresponding responses are sent back from the higher privileged code.
[0013] The present technology provides an alternative approach in which an interface (also referred to as a management interface in the embodiments described below) is provided in hardware between the processing circuitry and the virtualization device. The processing circuitry is configured to send a configuration request from the first software to the interface, requesting configuration of the virtualization device. The interface receives the request and responds by forwarding a mediation request to the processing circuitry, requesting the second software (at a higher privilege level) to determine a response to the configuration request received from the first software. The mediation request may simply be a request to perform mediation of the configuration request, may include the configuration request itself, or may include a modified version of the configuration request. In either case, the mediation request is a request that is subject to mediation, triggering the first software to determine a response to the configuration request.
[0014] Thereby, since the first and second software do not communicate directly with each other, they do not require a specific driver or a specific pre-agreed protocol to enable the mediation to be performed. Since the first and second software each only need to communicate with the interface according to an architecturally defined configuration command protocol, they may be compatible with other software using the same command architecture supported by the interface hardware. Thus, there is no need for direct collaboration between developers of higher and lower privilege code, or for the lower privilege code to be updated when ported to a different physical. In this way, the mediation of the configuration request is performed transparently, from the point of view of the first software, which sends a configuration request to the interface and receives a configuration response. The involvement of the second software is not highlighted to the first software.
[0015] In some embodiments of the present technology, after receiving the mediation request, the second software is configured to process the mediation request and send a mediation response to the interface in relation to the configuration request. The mediation response indicates a result of the mediation, such as whether the second software determined that the configuration request is allowed, or an indication of the result of the functionality emulated in the software of the virtualization device. In response to receiving the mediation response, the interface is configured to send a configuration response to the processing circuit based on the mediation response. The configuration response is thus a response to the configuration request from the first software.
[0016] This allows the mediation of configuration requests by the first (less privileged) software to be performed through an interface, avoiding the need for specific drivers for software applications to communicate directly with each other using memory access to shared data structures, making the system much more versatile.
[0017] In some embodiments, based on the brokered response, the interface is configured to trigger a virtualization device to be configured as requested by the first software.
[0018] This also saves processing time by eliminating the need for the first software to first request mediation through the higher privileged software and then request configuration if the mediation is successful. Instead, the configuration request is sent directly to the interface, and if the mediation is successful, the configuration request may be processed without the need to issue a second request. When the interface receives a configuration response from the processing circuitry, the mediation response may trigger configuration of the virtualization device based on the configuration request if it indicates that the configuration request is permitted.
[0019] The configuration response sent by the interface to the processing circuit can take many forms, depending in part on the mediated response from the second software. In some embodiments, if the mediated response indicates that the configuration request should be rejected, the configuration response sent to the first software includes an error indication. If an error indication is received from the hardware interface in response to the configuration request, the first software may fall back to other forms of mediation, e.g., by direct interaction with higher privileged software, or may abandon the request to configure the device.
[0020] In other embodiments, the mediated response may provide an emulated response to the configuration request that is generated in software by the second software, rather than the configuration request being processed directly by the virtualization device. In this case, the configuration response sent to the first software may be the same as the mediated response provided by the second software. That is, the second software may determine an appropriate response to the configuration request and provide the generated response to the configuration request as the mediated response, which is forwarded by the interface directly to the second software without invoking any function in the virtualization device. This is useful for handling certain configuration features that may not be supported by a particular model of the virtualization device.
[0021] In some embodiments of the present technology, the processing circuitry limits the issuance of mediated responses to software applications above a certain threshold privilege level.
[0022] This may mean that the processing circuitry ensures that the mediated request is specifically targeted to software with a sufficiently high privilege level (e.g. the mediated request could occur by triggering an exception that is handled at the threshold privilege level, or by updating data in a mailbox address that is only accessible to applications at or above the threshold privilege level), or it may mean that the privilege level of the second software is checked to see if it is sufficiently high before any mediated response is allowed to be issued (attempting to issue a mediated response that would cause a failure or be ignored in a state less privileged than the threshold privilege level). If the privilege level of the second software is not sufficiently high, the processing circuitry may send an error indication to the interface.
[0023] The advantage of this is that only software that has a sufficient level of knowledge about other applications that can access the virtualized device can perform the intermediation.
[0024] In some circumstances, it may be more efficient not to mediate any configuration requests. Thus, in some embodiments, an interface may be configured in one of a number of states. For example, these may include a mediation enabled state and a mediation disabled state. In the mediation enabled state, the interface responds to configuration requests by forwarding corresponding requests to the processing circuitry, as described above. On the other hand, in the mediation disabled state, upon receiving a configuration request, the interface does not forward the mediation request to the processing circuitry. In other words, in the mediation disabled state, forwarding of mediation requests is suppressed or omitted.
[0025] In some embodiments, the interface can accept configuration requests made when the intermediation disabled state is set. This can be useful in scenarios where there is only one application / user configured to use the virtualized device, there is no risk of interference with other applications / users, so no intermediation is required and configuration requests by the single application / user can be accepted. Additionally, the intermediation disabled state can be set if the virtualized device itself has the capability to mediate requests.
[0026] In some embodiments, if the mediation disabled state is set, the interface is configured to send an error indication to the processing circuit in response to receiving a configuration request. This may be useful in systems that run legacy code at a higher privilege level that is not written to be compatible with architecturally defined mediation request / response commands. In this case, returning an error indication can inform the first software at a lower privilege level that mediation is not possible at the architectural level, and the first software can fall back to other ways of requesting mediation, for example by interacting with the second software via a pre-agreed protocol based on a shared data structure in memory.
[0027] The mediation enabled or disabled state may be identified at the interface in a variety of ways: In some embodiments, the interface includes a register for sorting the enable indicator to indicate whether the mediation enabled or disabled state is set.
[0028] The enable indicator can take any form, but in some embodiments may be a single bit, for example, where a logic "1" value indicates an intermediation enabled state and a logic "0" value indicates an intermediation disabled state, or vice versa. In either case, storing the enable indicator in a register in the interface is an efficient way to indicate the intermediation enabled / disabled state.
[0029] Setting the mediation enabled or mediation disabled state by the processing circuitry may, in some embodiments, be accomplished by the processing circuitry sending a mediation set command to the interface.
[0030] The mediation set command instructs the interface to toggle between a mediation enabled state and a mediation disabled state. This allows software executing on the processing circuit, e.g., a second software of higher privilege, to control whether mediation is enabled. For compatibility with legacy code, the default state (when no mediation set command is received) may be for the interface to be in the mediation disabled state.
[0031] In some embodiments, if the interface includes a register that stores the enable indicator, the mediated set command triggers the interface to change the value of the enable indicator stored in the register.
[0032] For example, if a value of "1" indicates that mediation is enabled, then the interface is configured to respond to a set mediation command by writing a "0" instead of a "1." This is an efficient and easy to implement way to set the mediation state, and only requires a single command type to be able to toggle the mediation enabled / disabled state in either direction.
[0033] Alternatively, dedicated enable and disable commands can be defined to respectively enable and disable intermediation at the interface.
[0034] In some embodiments, the processing circuitry is configured to restrict issuance of mediated configuration commands to software at or above a given privilege level.
[0035] This increases the security of the system by ensuring that only software with a particular set of privileges can modify the mediated state of an interface.
[0036] In some embodiments, the mediation enabled or disabled state may be set individually for each of a number of different types of configuration commands, and in response to a particular type of configuration command, the interface is configured to determine whether to forward the mediation request to the processing circuitry depending on whether the mediation enabled or disabled state is set for the particular type of configuration command.
[0037] Ensuring that the mediation enable / disable state is set only for certain types of configuration commands improves overall system efficiency by preventing waste of power and processing time by rejecting configuration commands of types that could enable mediation (and vice versa). This approach allows the mediation state to be better tailored to the particular system in which the device is used. For example, some configuration commands may relate to device-specific features that are supported in some models of the virtualization device but not in other models of the virtualization device. Configuration commands that relate to features that are directly supported in the virtualization device's hardware can be set to disable mediation, while configuration commands that relate to features that are not supported in the virtualization device's hardware can be set to enable mediation, allowing the corresponding features to be emulated in software of a higher-privileged second piece of software in response to a corresponding mediation request.
[0038] In some embodiments of the present technology, the configuration command is associated with a device context identification value that indicates an associated device context that includes at least one parameter associated with the virtualized device when used by the first software.
[0039] In the apparatus of the present technology, there may be many software applications utilizing the same virtualized device, each with different requirements. Thus, each has its own device context, specifying one or more parameters of the virtualized device, and associated with an identification value. This is useful for recognizing the source of configuration requests received by the interface, for example, to allow the second software or interface to manage quality of service criteria that govern a fair share of resources in the virtualized device among competing applications seeking to use the device. It is also useful in some embodiments to be able to identify when the interface receives a particular request based on the identity of the first software, without having to mediate a particular request from a particular software application. Providing a device context identification value helps to enable this.
[0040] There are several types of configuration requests that can be sent from the processing circuitry to the interface, some examples of which are given below. It will be appreciated that these are only a few examples of the types of configuration commands to which the present technology is applicable, and that many other types may be equally suitable. One example is a request related to power management of resources of a virtualized device.
[0041] A software application with a lower privilege level (e.g., the first software) may not be aware of other software that uses the resources of the virtualized device. This means that the first software may believe that it is safe to instruct the device or part of the device to power down once the first software is no longer using the device. However, there may be other software applications using the device that the first software does not know about, so powering down the device or part of the device will have a negative impact on the other software. Therefore, it is useful for such power management commands to be mediated by a higher privilege software (e.g., the second software) that is aware of all the software that utilizes the virtualized device. In this way, it is beneficial to apply the present technology to configuration commands related to power management.
[0042] Another example of the type of configuration request to which this technique can be applied is a request that is not supported by the hardware.
[0043] Such a request may be a request to configure or use a resource not present in the virtualization device, or a request that the virtualization device perform an operation that the virtualization device does not actually support in hardware. In this case, the more privileged software (e.g., the second software) may emulate the virtualization device's response to the configuration command. Thus, the mediation request may trigger processing circuitry to execute the second software to emulate the unsupported operation requested to be performed by the virtualization device. By providing an architecturally defined framework of configuration / mediation commands and responses, and a hardware interface for forwarding these commands between the first and second software, the software emulation can be made transparent to the first software, so that the first software believes that the configuration is performed by the virtualization device itself.
[0044] A further example of the type of configuration request to which this technique can be applied is a request from an unallocated device context.
[0045] When a configuration request from software with an unallocated device context is received by the interface, the interface may mediate the configuration request to determine whether there are resources available to allocate to the device context, or whether the unallocated device context must wait for the resources to be released from use by another device context. Thus, again in this case the present technique is particularly useful.
[0046] Virtualization devices can take several possible forms, including hardware accelerators and input / output devices, although it will be appreciated that many other examples are possible.
[0047] A hardware accelerator is specialized computer hardware designed to perform a particular function more efficiently than software running on a general-purpose CPU. Examples of hardware accelerators include artificial intelligence (AI) accelerators such as graphics processing units (GPUs), neural processing units (NPUs), or cryptographic accelerators. This is a particularly beneficial situation for applying the present technology because both the hardware accelerator and input / output devices, such as input ports, network interfaces, or other types of input / output devices, can be shared by many software applications.
[0048] In some embodiments, the processing circuitry is configured to send a device action trigger command from the first software to the interface, and the interface is configured to send the device action trigger command to the virtualization device, omitting sending an intermediation request to the processing circuitry in response to the device action trigger command.
[0049] Device action triggering commands are commands that cause a virtualization device to perform a particular action, such as processing data. These commands can be sent directly to the virtualization device without any intermediaries, as the virtualization device is configured to do so. In this way, processing time is saved by avoiding unnecessary intermediaries.
[0050] For example, if the virtualization device is a GPU or other hardware accelerator, a direct device action trigger command may request processing of data by the hardware accelerator, while a configuration command may be a request to configure a share of processing resources available to the application issuing the command, a request to change configuration settings of the accelerator, or a request to change the power mode of the accelerator, etc.
[0051] In some embodiments, the virtualization device may not support a particular device action triggering command, in which case the interface may be configured to forward further mediation requests to the second software, which may then mediate the device action triggering command and emulate the virtualization device to determine a response to the device action triggering command.
[0052] This makes it possible to process commands that are not supported by the hardware as if they were supported, thereby improving the versatility of the system.
[0053] In some embodiments, the technology includes a computer program that controls a host data processing device to provide an environment for executing instructions by the device described in any of the above examples. The computer program may be a simulation program that includes processing program logic and interface program logic to cause the host processing system to perform some of the functions of the processing circuitry (executing software) and interface (receiving configuration requests from a first software executing in the processing program logic and forwarding the mediation requests to a second software executing in the processing program logic). The simulator program, when executed on the host data processing device, can emulate the functions of these hardware components so that other applications (designed to run on a hardware device having these components) that run on top of the simulator program can continue to operate in the same way even when executed on a hardware platform that does not have some or all of these hardware components.
[0054] The present technology relates to a system in which multiple processes operate at different privilege levels. Before describing the details of the technology, an overview of the concept of privilege levels will be provided by way of an example.
[0055] Fig. 1 shows a schematic example of privilege levels (exception levels) of processes that can access the virtualization device. Running at privilege level EL2 (exception level 2) is the hypervisor 102. The hypervisor 102 may be implemented as a set of instructions executed by a processing circuit, as a dedicated hardware component, or in software. The hypervisor 102 manages the execution of the virtual machines 104 and controls their access to the resources of the virtualization device. The hypervisor 102 with the highest privilege level is aware of and has access to all running virtual machines 104 and allows it to mediate any command sent from the virtual machines 104 to the virtualization device.
[0056] The virtual machines 104 run at the next privilege level, EL1. Each virtual machine 104 is unaware of the operation of the other virtual machines because they all run at the same exception level. This means that one of the virtual machines 104 may unwittingly issue a configuration request to a virtualization device that adversely affects one or more of the other virtual machines 104. Therefore, it may be useful for some configuration requests from the virtual machines 104 to be mediated by the hypervisor 102.
[0057] Multiple applications 106 run within each virtual machine 104, with the applications 106 running at the lowest privilege level, EL0. Each virtual machine 104 controls the applications 106 running within its guest operating system, for example, VM0 104a controls the first two applications 106a, 106b shown, but cannot control applications 106 running in other virtual machines 104.
[0058] Thus, each privilege level defines the level of control a process has over other processes running on the virtualization device, with processes running at EL2 controlling processes running at EL1 and EL0, and processes running at EL1 controlling processes running at EL0. Furthermore, each privilege level is linked to a particular set of permissions, with more privileged processes having greater access than less privileged processes. It will be appreciated that in some architectures more than three privilege levels may be defined. For example, a secure monitor running at a privilege level EL3, which is higher than the hypervisor's privilege level EL2, may control transitions between secure and non-secure domains. It will be appreciated that this is only one example of a possible hierarchy of privilege levels, and that other implementations may use a different set of privilege levels, but in general a processing system will support the execution of software processes at different privilege levels, where a more privileged process has the authority to perform actions or access data not permitted by a less privileged process.
[0059] FIG. 2 illustrates an example of a system 200 with multiple privilege levels, as illustrated in FIG. 1. In FIG. 2, two software processes 202, 204 have access to a virtualization device 206. The software 202, 204 execute on a processing circuit (not shown), which may be either the same processing circuit that executes both software processes 202, 204, or a separate set of processing circuits. The virtualization device 206 is a hardware component capable of performing operations under the control of the software processes 202, 204. For example, the virtualization device 206 may be an input / output device, or a hardware accelerator configured to perform certain functions more efficiently than software running on a general-purpose CPU, such as a dedicated graphics processing unit (GPU), an AI accelerator or neural processing unit (NPU) for use in artificial intelligence applications, or a cryptographic accelerator for performing cryptographic operations.
[0060] Each of the software processes 202, 204 has an assigned privilege level, one being higher privilege software 204 and the other being lower privilege software 202. For example, the lower privilege software 202 may be a virtual machine 104 as shown in FIG. 1 and the higher privilege software 204 may be a hypervisor 102 or a particular set of instructions executing within the hypervisor 102. Each of the higher and lower privilege software 202, 204 may wish to send a request to configure the virtualization device 206 to receive data processing instructions from that software. The configuration request from the lower privilege software 202 needs to be mediated by the higher privilege software 204 because it has a lower privilege level.
[0061] 2, mediation is performed by communication between lower privilege software 202 and higher privilege software 204. Lower privilege software 202 sends a mediation request to higher privilege software 204 in response to higher privilege software 204 determining whether to allow or deny the configuration request from lower privilege software 202 based on knowledge of other software that can access resources of virtualization device 206. Higher privilege software 204 then returns a mediation response to lower privilege software 202 indicating whether the configuration request is allowed. If the request is allowed, lower privilege software 202 sends the configuration request to virtualization device 206.
[0062] Such configuration requests and responses may include, for example, writing to data structures in memory according to an agreed upon application-specific protocol. Such writing would require the cooperation of software developers of both higher-privilege software 204 and lower-privilege software 202 in a manner custom-tailored for a particular application. In this manner, if multiple lower-privilege software applications are to interact with the same device, the developer of higher-privilege software 204 may agree to different data structures and protocols for interacting with each item of lower-privilege software.
[0063] As shown in FIG. 2, there are many advantages to requiring mediation of configuration commands from the lower privilege software 202. For example, mediation reduces the possibility of problems caused by contention for resources of the virtualization device 206, improving the security of the system 200. However, from the perspective of system efficiency and versatility, this particular approach also has significant drawbacks. One drawback is, for example, that each configuration request from the lower privilege software 202 needs to be mediated before being sent to the virtualization device 206. This increases the processing time it takes for the request to reach the virtualization device, reducing the efficiency of the system 200. Another drawback is that, for mediation to be performed, the two software processes 202, 204 need to communicate directly. This requires that each software process 202, 204 includes specific drivers 203, 205 to enable communication with the other software process. These software drivers 203, 205 need to be updated or reinstalled every time a software process is replaced, which means that the versatility of the system 200 is limited. This is particularly problematic in modern systems where multiple users share access to a single server and where users may switch from one server to another as prices or other conditions change. It would therefore be beneficial to improve the efficiency and versatility of system 200.
[0064] FIG. 3 illustrates a system 300 using an example of the present technology. In the system illustrated in FIG. 3, a management interface 302 is provided between the software 202, 204 and the virtualization device 206. In this configuration, the lower privilege software 202 is configured to send a configuration request to the management interface 302 without first requesting mediation by the higher privilege software 204. The management interface 302 is then responsible for sending a mediation request to the higher privilege software 204 for the configuration request. Although the mediation of the configuration request is performed by the higher privilege software 204, the mediation response is returned to the management interface 302, not to the lower privilege software 202. If the mediation response indicates that the configuration request is accepted, the management interface 302 sends a configuration response indicating this to the lower privilege software 202, triggering the virtualization device 206 to be configured according to the configuration request. On the other hand, if the mediated response indicates that the configuration request is rejected, then the management interface 302 sends a configuration response to the less-privileged software 202 to indicate this, for example in the form of an error indication.
[0065] In the above example, the mediated response indicates whether the configuration request can be accepted or rejected.
[0066] However, in other examples, the mediated response may include an emulated response to the configuration request that may be sent to the interface 302 and forwarded to the first software 202 without requiring any action at the virtualization device 206. For example, this may be useful for handling certain configuration actions that are not actually supported in hardware by a particular model of the virtualization device.
[0067] Configuration requests and responses, and mediation requests and responses may be architecturally defined such that hardware devices conforming to a particular architecture ensure that software designed according to the architecture operates in a predictable manner. As a result, higher and lower privilege software 204, 202 can be designed to use configuration and mediation requests and responses in the manner described above without having to agree on a mediation protocol with other software (the agreement on the protocol is effectively between each software provider and a hardware provider or standard, rather than between two software providers, which is more efficient for software development).
[0068] This is an example of transparent mediation of configuration requests, where from the perspective of the lower-privilege software 202, the mediation of the configuration request is hidden from the lower-privilege software 202, such that the configuration request is sent to the management interface 302 (it is even possible to program the lower-privilege software 202 as if it were sending the configuration request directly to the virtualization device 206) and a response is received from the management interface 302. Thus, the mediation is performed without the involvement of the lower-privilege software 202, and without the lower-privilege software 202 needing to be aware of the involvement of the higher-privilege software 204.
[0069] In this arrangement, since the mediation request is sent by the management interface 302 and not the lower privilege software 202, the dedicated driver 203 for communicating with the higher privilege software 204 does not need to be provided in the lower privilege software 202, and the dedicated driver 205 for communicating with the lower privilege software 202 does not need to be provided in the higher privilege software 204. This improves the versatility of the system 300, allowing either of the software programs 202, 204 to be replaced without having to update or reinstall the drivers 203, 205 of the other software. Furthermore, this arrangement allows the lower privilege software 202 to send configuration requests directly to the management interface 302 without waiting for the higher privilege software 204 to mediate the request first. In this manner, the processing time required to send a configuration request is reduced, improving overall efficiency.
[0070] The arrangement shown in Figure 3 has uses other than mediating configuration requests. For example, while most device action triggering commands (described in detail with reference to Figures 13-15) do not need to be mediate, device action triggering commands that are not supported by the virtualization device 206 may be mediate, which may be performed according to the techniques described above.
[0071] Furthermore, the above mechanism not only serves to broker requests, but can also be used to enable any other communication between the first and second software, allowing communication between software applications without the need for dedicated drivers.
[0072] FIG. 4 shows a more detailed example of a hardware system 400 to which the present technology can be applied. In FIG. 4, several examples of virtualization devices are shown, including a network interface 402, a cryptographic unit 404, and a graphics processing unit (GPU) 406. Each of the virtualization devices 402, 404, 406 is an example of the virtualization device 206 shown in FIG. 3, and can access one or more of a plurality of processing units 408, 410, and all of the processing units 408, 410 and the virtualization devices 402, 404, 406 can access a shared memory 412. The GPU 406 includes several components, including a processing circuit 414 for performing operations under the control of one or more software applications executing on the processing units 408, 410. The processing circuit can access a cache 416 for storing data or instructions retrieved from the memory 412, and a register bank 418. Also provided within the GPU 406 is a management interface 420. 3, and is configured to receive configuration requests and send arbitration requests from processing units 408, 410. In this example, management interface 420 is shown as a component of GPU 406, although it is possible for management interface 420 to be a separate component to the virtualization device, as illustrated by management interfaces 422, 424 provided to network interface 402 and cryptographic unit 404, respectively. Management interface 420 also has access to register bank 418.
[0073] Each of the processing units 408, 410 includes processing circuitry 426, 428 configured to perform operations to process instructions. The instructions processed by the processing circuitry 426, 428 may represent one or more software applications, such as the software applications 202, 204 shown in Figure 3, each having an established privilege level. The processing circuitry 426 has access to register banks 430, 432 and caches 434, 436 configured to store data or instructions retrieved from the memory 412.
[0074] When a first software executing on a processing circuit 426 of one of the processing units 408 requests configuration of one of the virtualization devices, for example the GPU 406, the processing circuit 426 is configured to send a configuration request to a management interface 420 associated with the GPU 406. The management interface responds to the configuration request by forwarding the mediation request to a processing circuit executing a second software of higher privilege, which may be the same processing circuit 426 that executes the first software or a different processing circuit 428. The processing circuit 426, 428 that receives the mediation request is configured to mediate the configuration request and send a mediation response to the management interface 420, which is configured to send the configuration response to the processing circuit 426 that executes the first software and selectively trigger configuration of the GPU based on the configuration request.
[0075] FIG. 5 shows an overview of a system 500 according to an example of the present technology. The system includes a virtualization device 206 and a processing circuit 502 that communicate through a management interface 302. A first software 504 and a second software 506 with different privilege levels, which are examples of the lower privilege and higher privilege software 202, 204 shown in FIG. 3, run on the processing software 502. Since the second software 506 has a higher privilege level than the first software 504, when the first software 504 sends a configuration request to the management interface 302, an intermediation request is sent from the management interface 302 to the second software 506. The management interface 302 receives an intermediation response from the second software 506 and sends an appropriate configuration response back to the first software 504. If the intermediation indicates that the configuration request is allowed, the management interface 302 also triggers the virtualization device 206 to be configured according to the configuration request.
[0076] In this example, the first and second software 504, 506 are shown executing on the same processing circuitry 502. However, it will be understood that this is only one possible configuration and that the first and second software 504, 506 may also execute on separate sets of processing circuits.
[0077] FIG. 6 is a flow diagram illustrating an example of a method 600 performed by an administrative interface (such as administrative interfaces 302, 420, 422, 424 shown in FIGS. 3-5) in accordance with an example of the present technology.
[0078] In a first step S602, the management interface is configured to receive a configuration request from a processing circuit 426, 428, 502 executing the first software 202, 504. The configuration request is a request to configure a virtualization device associated with the management interface. In response to receiving the configuration request, the management interface forwards a mediation request to a processing circuit executing the second software 204, 506, S604. The mediation request is a request to mediate the configuration request from the first software. Then, in response to sending the mediation request, the management interface is configured to receive a mediation response from the second software, S606, indicating whether the configuration request is allowed. Then, the management interface sends a configuration response to the processing circuit executing the first software, S608, indicating whether the configuration request is accepted. Finally, following the determination of whether the configuration request is accepted at S610, the process either ends at S612 without configuring the virtualization device (if the configuration request was not accepted) or ends after triggering configuration of the virtualization device in accordance with the configuration request at S614.
[0079] 7A and 7B are timing diagrams showing some of the signals passing between components of the system 300, 400, 500 according to the present technology. In FIG. 7A, a configuration request 702 is first sent from the first software 704 to the interface 706. Then, a mediation request 708 is sent from the interface to the second software 710, followed by a mediation response 712 sent from the second software 710 to the interface 706. Then, a configuration response 714 is sent from the interface to the first software. This demonstrates the invisibility of the mediation of the configuration request by the second software 710 to the first software 704. Looking at the signals sent and received by the first software, only the configuration response 714 received in response to the configuration request 702 is seen. Since there is no direct communication between the first software 704 and the second software 710, no dedicated driver or protocol is required to communicate with the second software. The first software 704 is an example of the lower privilege software 202, 504 shown in Figures 3 and 5, the second software 710 is an example of the higher privilege software 204, 502 shown in Figures 3 and 5, and the interface 706 is an example of the management interface 302, 420, 422, 424 shown in Figures 3-5.
[0080] 7B, on the other hand, shows signals passing between a processing circuit 720 executing the first and second software, an interface 706, and a virtualization device 722, where the processing circuit 720 is an example of the processing circuit 426, 428, 502 shown in FIGS. 4 and 5, and the virtualization device 722 is an example of the virtualization device 206, 402, 404, 406 shown in FIGS. 3-5. In FIG. 7B, a configuration request 702 is sent from the processing circuit 720 to the interface 706, which responds by sending a mediation request 708 to the processing circuit 720, requesting mediation of the configuration request 702 by the higher privileged software. A mediation response 712 is sent by the processing circuit 720 to the interface 706, which returns a configuration response 714 to the processing circuit 720. Concurrent with sending the configuration response 714 to the processing circuit, the interface 706 sends a configuration command 724 to the virtualization device 722 based on the configuration request 702, which triggers the virtualization device 722 to become configured according to the configuration request 702. Note that although the configuration command 724 and the configuration response 714 are shown in FIG. 7B as occurring at the same time, it is possible for them to occur at slightly different times, so long as neither occurs before the mediated response 712 is received from the interface.
[0081] In some embodiments of the present technology, an intermediation enabled state or an intermediation disabled state can be set. In the intermediation enabled state, the mediation of the configuration request according to the technology described above is easy. In the intermediation disabled state, the mediation of the configuration request is not performed. This is useful in situations where the result of the mediation of the configuration request is likely to always be the same. For example, if the first software is the only software accessing a particular resource on the virtualization device, the configuration request may always be granted. Conversely, if all resources are currently being used by software with higher privileges than the first software, or if there is some fault, the configuration request may always be rejected. In either case, the result of the mediation is always the same, so there is no need to mediate the configuration request. In this way, it is beneficial to switch to the intermediation disabled state in order to save processing time and power by preventing unnecessary mediation from occurring. However, in other cases, the mediation is still required, so the intermediation enabled state needs to be set.
[0082] The mediation enabled or disabled state may be set in a management interface in a variety of ways. FIG. 8 illustrates an example of a management interface 802 according to the present technology, including a register 804, which is an example of the management interfaces 302, 420, 422, 424, 706 of FIGS. 3-5, 7A, and 7B. The register 804 includes an entry that stores an enable indicator that indicates whether the mediation enabled state or the mediation disabled state is set. The enable indicator may be in the form of a single bit (e.g., a logical "1" value indicates a mediation enabled state and a logical "0" value indicates a mediation disabled state, or vice versa) or in other forms. The mediation enabled or disabled state may be set, for example, by changing the enable indicator in the register in response to a mediation setting command from the processing circuitry.
[0083] Registers 804 may be provided within management interface 802, as shown in FIG. 8, or may be registers within a register bank within a virtualization device, such as register bank 418 in GPU 406 of FIG.
[0084] The management interface 802 shown in FIG. 8 also includes decision logic 806 configured to determine, in response to a configuration request from the processing circuit executing the first software, whether to send a mediation request to the processing circuit executing the second software based on the enable indicator set in the register 804.
[0085] In some embodiments, rather than setting the mediation enabled or mediation disabled state for every configuration request received by the management interface, the state may be set individually for several different types of configuration commands. FIG. 9 illustrates an example of a possible register 902. The register 902 has eight entries 904, each for a different type of configuration request. Each entry stores an enable indicator that indicates whether the mediation enabled or mediation disabled state is set for the corresponding type of configuration command. In this example, each enable indicator is a single bit value, with a "1" representing a mediation enabled state for that type of configuration request and a "0" representing a mediation disabled state for that type of configuration request. For example, the enable indicators for entries 904b-d, 904f-g for configuration request types BD and FG are "1", indicating that the mediation enabled state is set for each of these types. This means that when a configuration request of any of these types is received by the management interface, the mediation request is forwarded to the processing circuitry according to the above example. On the other hand, the enable indicators for entries 904a, 904e, 904h for configuration request types A, E, and H are set to "0" to indicate that the mediation disabled state is set for these types, meaning that when a configuration request of any of these types is received by the management interface, the mediation request is not forwarded to the processing circuitry.
[0086] It will be appreciated that although the register 902 shown in FIG. 9 shows only eight entries 904 for eight types of configuration requests, it may include any number of entries 904. The types of configuration requests referenced may include, for example, configuration requests related to power management of resources on a virtualized device, configuration requests that are not supported by hardware, and configuration requests from an unallocated device context. However, these are only one example of types of configuration requests. Other types of configuration requests are possible. Additionally, the register 902 is only one example of the register 804 shown in FIG. 8. Also, in some cases, multiple types of requests may share a single indicator 904, such that each of those types are simultaneously enabled / disabled by the setting of the shared indicator, and other types may include separate enable / disable indicators 904.
[0087] FIG. 10 is a flow diagram illustrating an example of a method 1000 performed by a management interface according to an example of the present technology. The method begins with step S1002 of receiving a configuration request of a given type from a processing circuit executing a first software. Then, in S1004, a determination is made, for example by the decision logic 806 shown in FIG. 8, as to whether mediation is enabled for the configuration request of the given type. For example, this may include the decision logic 806 reading an enable indicator 906 stored in an entry 904 of the register 902 corresponding to the configuration request of that type. If it is determined that mediation is disabled for the configuration request of the given type, the method moves to step S1006 of sending an error indication to the processing circuit executing the first software as a configuration response. This indicates to the first software that the configuration request has been rejected. The first software can then fall back to another way of proceeding by emulating the functionality of the virtualization device itself in software.
[0088] On the other hand, if it is determined that mediation is enabled for the given type of configuration request, the method proceeds to step S1008 of forwarding the mediation request to a processing circuit executing a second software having a higher privilege level than the first software. In response to the mediation request, in step S1010, a mediation response is received from the processing circuit executing the second software indicating whether the configuration request is accepted or rejected. In step S1012, a determination is made, for example by the decision logic 806, as to whether the mediation response indicates that the configuration request is accepted. If it is determined that the configuration request is rejected, in step S1006, an error indication is sent to the processing circuit executing the first software as a configuration response. On the other hand, if the configuration request is accepted, the method proceeds to step S1014 of sending an indication forming a configuration request to the processing circuit executing the first software indicating that the configuration request is accepted. And in step S1016, based on the configuration request, the interface triggers configuration of the virtualization device.
[0089] 11 illustrates an example of a method 1100 performed by a processing circuit according to an example of the present technology. The method begins with step S1102 of sending a configuration request from first software executing on the processing circuit to an interface. In response, at S1104, a mediation request is received from the interface. At S1106, the mediation request is processed by second software executing on the processing circuit, the second software having a higher privilege level than the first software, and the mediation request is processed by mediating the configuration request. Once the mediation is completed, the method proceeds to a final step S1108 of sending a mediation response to the interface indicating whether the configuration request is accepted or rejected.
[0090] 12A and 12B are flow diagrams illustrating methods involved in setting an intermediation enabled or disabled state performed by the processing circuit and the interface, respectively.
[0091] 12A shows that if it is determined in S1202 that software of a given privilege level or higher is running, then in S1204 an intermediated configuration command is sent by the processing circuitry to the interface, to ensure that system security is maintained.
[0092] FIG. 12B shows that when the interface receives an mediated setting command in S1206, the enable indicator stored in the register is changed in S1208.
[0093] 13 shows an example of the present technology in a schematic manner. The figure shows one or more devices 1302 operable as virtualized devices under the control of a virtualization application 1304. The virtualization application 1304 (an example of the lower privilege first software 202, 504, 704 described in the previous example) includes a CP (control path) driver 1306 configured to send control information such as configuration requests to the device 1302 via a virtualization device interface 1308, and a DP (data path) driver 1310 configured to send commands (device action trigger commands) to the device 1302 to trigger certain processing operations to be executed on the device 1302.
[0094] When a configuration request is sent from the virtualization application 1304 to the virtualization device interface 1308 via the CP driver 1306, the interface 1308 responds by forwarding the mediated request to the privileged CP driver 1314 via the device interface 1312. The privileged CP driver 1314 returns a mediated response to the virtualization device interface 1308 via the device interface 1312, and the virtualization device interface 1308 sends a corresponding configuration response to the virtualization application 1304 via the CP driver 1306. The virtualization device interface 1308 is an example of the management interface 302, 420, 422, 424, 706, 802 as shown in the previous example.
[0095] A device action trigger command sent via the virtualization application's DP driver 1310 or by another DP driver 1316 associated with the device 1302 does not usually need to be mediated and is passed to the device 1302 by the virtualization device interface 1308 and the device interface 1312, respectively. However, in some cases, mediation is required. For example, if the device action trigger command is not supported by the virtualization device's hardware, the device interface 1312 can be configured to forward further mediation requests to second software running on the processing circuitry. The second software can then mediate the device action trigger command and emulate the virtualization device's response to the device action trigger command.
[0096] Figure 14 illustrates an example of a situation in which application of the present technology is particularly advantageous. In particular, Figure 14 illustrates the application of the present technology to over-provisioning of device interfaces.
[0097] In FIG. 14, the virtualized device interface 1308 is shared between the virtualized application 1304 shown in FIG. 13 and an additional virtualized application 1404. The additional virtualized application 1404 is deployed in the same manner as the first virtualized application 1304, along with the DP driver 1410 and the CP driver 1406. In this example, the virtualized device interface 1308 is over-provisioned, i.e., some of its resources are assigned to both the first virtualized application 1304 and the second virtualized application 1404, and both believe that they are the only applications that can access those resources. Traditionally, in this situation, any attempt to access the over-provisioned resource would require trapping 1412 of memory accesses on both the control path 1402 and the data path 1408. However, according to the present technology, memory accesses on the control path 1402 do not need to be trapped, because they are transparently mediated through the privileged CP driver 1314 by the software or hardware responsible for managing the over-provisioned resource.
[0098] Another example of a situation in which the present technology is particularly advantageous is live migration, as shown in Fig. 15. In a live migration of a virtualized application 1304 from one host device 1302 to another host device 1502, a higher privilege software pre-copies the assigned context from the first device 1302 to the second device 1502 while ensuring that the virtualized application 1304 does not further change the configuration of the device 1302 without authorization. Conventionally, this requires specific operations to be programmed into the virtualized application 1304 or the CP driver 1306, but the transparent mediation of configuration commands as described according to the present technology means that this is no longer required.
[0099] Furthermore, the second device 1502 may not support the exact same feature set as the first device 1302, which would traditionally prevent live migration from being performed successfully. However, the present technology allows the missing functionality of the second device 1502 to be emulated by higher privileged software, allowing live migration to be performed on a wider range of systems. In this way, the present technology improves both the security and versatility of the system.
[0100] Particular examples of situations in which the application of the present technique is particularly beneficial are shown in Figures 14 and 15, however it will be appreciated that these are merely examples and that there may be many other situations in which the present technique would provide advantages.
[0101] FIG. 16 illustrates a simulator implementation that may be used. While the previously described embodiments implement the invention in terms of apparatus and methods for operating specific processing hardware supporting the technology, it is also possible to provide an instruction execution environment according to the embodiments described herein that is implemented using a computer program. Such computer programs are often referred to as simulators insofar as they provide a software-based implementation of a hardware architecture. Types of simulator computer programs include emulators, virtual machines, models, and binary translators, including dynamic binary translators. Typically, a simulator implementation may run on a host processor 1630 that optionally runs a host operating system 1620 that supports a simulator program 1610. In some configurations, there may be multiple simulation layers between the hardware and the instruction execution environment provided, and / or there may be multiple different instruction execution environments provided on the same host processor. Historically, powerful processors have been required to provide simulator implementations that run at reasonable speeds, but in certain situations, such an approach may be justified, such as when one wants to run code native to another processor for compatibility or reuse reasons. For example, a simulator implementation may provide an instruction execution environment that has additional features not supported by the host processor hardware, or that are typically associated with a different hardware architecture. An overview of simulation is given in "Some Efficient Architecture Simulation Techniques", Robert Bedichek, Winter 1990 USENIX Conference, pp. 53-63.
[0102] While embodiments have been described above with reference to particular hardware configurations or functions, in the simulated embodiments, equivalent functionality may be provided by appropriate software configurations or functions. For example, particular circuits may be implemented as computer program logic in the simulated embodiments. Similarly, memory hardware such as registers or caches may be implemented in the simulated embodiments as software data structures. In configurations in which one or more of the hardware elements referenced in the embodiments described above reside on host hardware (e.g., host processor 1630), some simulated embodiments may utilize the host hardware where appropriate.
[0103] The simulator program 1610 may be stored in a computer-readable storage medium (which may be a non-transitory medium) and provides a program interface (an instruction execution environment) to the target code 1600 (which may include applications, operating systems, and hypervisors) that is the same as the application program interface of the hardware architecture being modeled by the simulator program 1610. In this manner, the program instructions of the target code 1600 and the commands triggered by those instructions, including configuration requests and responses, mediated requests and responses, and the device action trigger commands described above, may be executed from within the instruction execution environment using the simulator program 1610, such that a host computer 1630 that does not actually have the hardware capabilities of the device 400 described above can emulate those capabilities.
[0104] In particular, the functionality of the above interfaces may be emulated by program logic within the simulator program 1610. If the target code 1600 includes instructions that require a configuration request to be sent to an interface, the program logic may be configured to emulate the functionality of the interface by forwarding the mediation request to a higher privileged software, or by emulating the second software itself by mediating the request itself. Thus, the configuration request and response, the mediation request and response, and the mediation process may all be emulated within the simulator code 1610. Thus, in some examples, the second software may actually be the simulator program 1610 itself. In other examples, there may be a second software separate from the simulator program 1610.
[0105] In this application, the term "configured to" is used to mean that an element of an apparatus has a configuration that can perform a defined operation. In this context, "configuration" refers to an arrangement or method of interconnection of hardware or software. For example, an apparatus may have dedicated hardware that provides the defined operation, or a processor or other processing device may be programmed to perform the function. "Configured to" does not imply that an apparatus element must be modified in any way to provide the defined operation.
[0106] Although exemplary embodiments of the present invention have been described in detail herein with reference to the accompanying drawings, it is to be understood that the invention is not limited to those precise embodiments, and various changes and modifications can be made by those skilled in the art without departing from the scope of the present invention as defined by the appended claims.
Claims
1. 1. An apparatus comprising: a processing circuit configured to execute software including first software and second software, the second software having a higher privilege level than the first software; a hardware implemented interface implemented in hardware, the hardware implemented interface configured to receive a configuration request from the processing circuitry from the first software requesting configuration of a virtualization device; In response to the configuration request, the hardware implemented interface is configured to forward an arbitration request to the processing circuitry for response determination by the second software; The apparatus, wherein the mediation request includes a request for the second software to determine a mediation response that indicates to the hardware implementation interface whether the virtualization device should allow or deny configuration of the virtualization device according to the configuration request received from the first software.
2. the hardware implemented interface is configured to receive a mediated response from the processing circuitry in response to transmitting the mediated request; the hardware implemented interface is configured to send a configuration response to the processing circuit based on the mediated response, the configuration response being a response to the configuration request.
2. The apparatus of claim 1.
3. the hardware implemented interface is configured to trigger the virtualization device to be configured based on the mediated response.
3. The apparatus of claim 2.
4. if the mediation response indicates that the configuration request from the first software should be rejected, the configuration response includes an error indication.
4. Apparatus according to claim 2 or 3.
5. the processing circuitry is configured to restrict issuance of the mediated response to software having a particular privilege level or higher.
5. Apparatus according to any one of claims 2 to 4.
6. in a mediation enabled state, the hardware implemented interface is configured to forward the mediation request to the processing circuitry in response to the configuration request; and in a mediation disabled state, the hardware implemented interface is configured to inhibit forwarding of the mediation request to the processing circuitry in response to the configuration request.
6. Apparatus according to any one of claims 1 to 5.
7. and in the intermediation disabled state, in response to receiving the configuration request, the hardware implemented interface is configured to send an error indication to the processing circuit.
7. The apparatus of claim 6.
8. a register configured to store an enable indicator indicating whether the mediation enabled state or the mediation disabled state is set.
8. Apparatus according to claim 6 or 7.
9. the processing circuitry is configured to set the mediation enabled state or the mediation disabled state by sending a mediation set command to the hardware implemented interface.
9. Apparatus according to any one of claims 6 to 8.
10. a register configured to store an enable indicator indicating whether the intermediation enabled state or the intermediation disabled state is set; in response to the mediated configuration command, the hardware implemented interface is configured to change the value of the enable indicator stored in the register.
10. The apparatus of claim 9.
11. the processing circuitry is configured to restrict issuance of the mediated configuration command to software at or above a given privilege level.
11. Apparatus according to claim 9 or 10.
12. the hardware implementation interface is configured to set the mediation enabled state or the mediation disabled state separately for at least two types of configuration requests; In response to a configuration request of a given type, the hardware implemented interface is configured to determine whether to forward the mediation request to the processing circuitry depending on whether the mediation enabled state or the mediation disabled state is set corresponding to the configuration request of the given type.
12. Apparatus according to any one of claims 6 to 11.
13. the configuration request is associated with a device context identification value indicating an associated device context that includes at least one parameter associated with the virtualization device when used by the first software; 13. Apparatus according to any one of claims 1 to 12.
14. The configuration request is: requirements related to power management of resources of the virtualized device; Requests that are not supported by the hardware and One of the requests is from an unallocated device context.
14. Apparatus according to any one of claims 1 to 13.
15. the processing circuitry is configured to send a device action trigger command from the first software to the hardware implemented interface; the hardware implemented interface is configured to send the device action trigger command to the virtualization device and omit sending a mediation request to the processing circuit in response to the device action trigger command.
15. Apparatus according to any one of claims 1 to 14.
16. if the device action triggering command is not supported by the virtualization device, the hardware implemented interface is configured to forward a further mediation request to the processing circuit in response to the device action triggering command; the further mediation request includes a request for the second software to emulate the virtualized device to determine a response to the device action trigger command.
16. The apparatus of claim 15.
17. 1. A method comprising: executing, on a processing circuit, software including a first software and a second software, the second software having a higher privilege level than the first software; receiving, at a hardware implemented interface implemented in hardware, a configuration request from the processing circuitry, the configuration request from the first software requesting configuration of a virtualization device; and in response to the configuration request, forwarding a mediation request from the hardware implemented interface to the processing circuitry for the second software to determine a response; The method, wherein the mediation request includes a request for the second software to determine a mediation response that indicates to the hardware implementation interface whether the virtualization device should allow or deny configuration of the virtualization device in accordance with the configuration request received from the first software.
18. 1. An apparatus comprising: means for executing software including a first software and a second software, the second software having a higher privilege level than the first software; means for receiving, from the means for executing software, a dedicated configuration request from the first software requesting configuration of a virtualization device; and means for forwarding an intermediation request to the means for executing software in response to the dedicated configuration request, for the second software to determine a response; The apparatus, wherein the mediated request includes a request that a second software having a higher privilege level than the first software determine a mediated response that indicates to a hardware implementation interface whether the virtualization device should allow or deny configuration of the virtualization device according to the configuration request received from the first software.
19. 1. A computer program for controlling a host data processing device to provide an instruction execution environment, comprising: a processor program logic configured to execute software including a first software and a second software, the second software having a higher privilege level than the first software; and interface program logic configured to receive a configuration request from the first software executed on the processing program logic, the configuration request requesting configuration of a virtualization device; In response to the configuration request, the interface program logic is configured to forward an intermediation request to the second software executed by the processing program logic for the second software to determine a response; the second software has a higher privilege level than the first software; The program, wherein the mediation request includes a request for the second software to determine a mediation response that indicates to hardware program logic whether the virtualization device should allow or deny configuration of the virtualization device according to the configuration request received from the first software.
20. 20. A non-transitory computer readable storage medium configured to store a computer program according to claim 19.
Citation Information
Patent Citations
Restricting access to computing resources
US10084784B1
Method, system, and apparatus for delegating control over the configuration of multi-tenant network devices
US10230709B1
System and method for secure execution of program code
US20060047958A1
Creation and Use of Virtual Device Drivers on a Serial Bus
US20100153089A1
Virtual secure mode for virtual machines
US20150082305A1
