PCIe devices and their operation methods

By managing the PCIe link and receiving functional mode control information, and setting multiple operating modes for secondary physical functions, the problems of high operating costs and low resource utilization of PCIe devices are solved, achieving more efficient resource management and cost reduction.

CN115203101BActive Publication Date: 2026-11-17SK HYNIX INC
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202210175921.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2021-04-13
Filing Date
2022-02-25
Publication Date
2026-11-17
Estimated Expiration
2042-02-25

AI Technical Summary

Technical Problem

In the existing technology, PCIe devices have high operating costs and are difficult to manage the operation modes of multiple physical functions effectively, resulting in low resource utilization.

Method used

A PCIe device and its operation method are adopted. By managing the PCIe link and receiving functional mode control information, the operation modes of multiple second physical functions are set to active or inactive modes, thereby realizing flexible management of multiple physical functions.

Benefits of technology

It reduces the operating costs of computing systems, improves resource utilization, and enhances the management efficiency of PCIe devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115203101B_ABST
    Figure CN115203101B_ABST
Patent Text Reader

Abstract

This disclosure relates to a PCIe device and a method of operating the same. The high-speed peripheral component interconnect (PCIe) device, which communicates with a host via a PCIe link, includes a first physical function, multiple second physical functions, and a function mode controller. The first physical function manages the PCIe link and receives function mode control information from the host. Each of the multiple second physical functions can be enabled or disabled according to a specific operating mode. Based on the function mode control information, the function mode controller sets the operating mode of the multiple second physical functions to one of an active mode or an inactive mode.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Cross-references to related applications

[0002] This application claims priority to Korean Patent Application No. 10-2021-0048073, filed on April 13, 2021, which is incorporated herein by reference in its entirety. Technical Field

[0003] This disclosure relates to an electronic device, and more specifically to a high-speed peripheral component interconnect (PCIe) device and a method of operating the same. Background Technology

[0004] High-speed peripheral component interconnect (PCIe) devices can define bus protocols for connecting input / output devices to host devices. High-speed PCI (PCIe) has a programming concept defined by the PCI standard and includes a physical communication layer defined as a high-speed serial interface.

[0005] Storage devices can store data in response to control by host devices such as computers or smartphones. Storage devices can include memory devices that store data and memory controllers that control the memory devices. Memory devices can be classified as volatile memory devices and non-volatile memory devices.

[0006] Volatile memory devices can only retain data when powered, and may lose stored data when power is unavailable. Types of volatile memory devices include static random access memory (SRAM), dynamic random access memory (DRAM), etc.

[0007] Non-volatile memory devices do not lose data even when there is no power supply. Types of non-volatile memory devices can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), and flash memory. Summary of the Invention

[0008] Various embodiments relate to a PCIe device and a method of operating the same that can reduce the operating costs of a computing system including a PCIe device.

[0009] According to an embodiment, a high-speed peripheral component interconnect (PCIe) device may include: a first physical function that manages the PCIe link and receives function mode control information from a host; a plurality of second physical functions that are enabled or disabled according to their respective operating modes; and a function mode controller that sets the operating modes of the plurality of second physical functions based on the function mode control information, each of the operating modes being set to one of an active mode and an inactive mode.

[0010] According to an embodiment, a method for operating a high-speed peripheral component interconnect (PCIe) device, the PCIe device including a first physical function and a plurality of second physical functions, the first physical function managing PCIe links connected to a host, the method may include: receiving function mode control information from a host through the first physical function; and setting an operating mode for each of the plurality of second physical functions based on the function mode control information, each operating mode being set to one of an active mode and an inactive mode. Attached Figure Description

[0011] Figure 1 An example of a computing system according to an embodiment of the present disclosure is shown;

[0012] Figure 2 Another example of a computing system according to an embodiment of the present disclosure is shown;

[0013] Figure 3 Another example of a computing system according to an embodiment of the present disclosure is shown;

[0014] Figure 4 A multifunctional device according to an embodiment of the present disclosure is shown.

[0015] Figure 5 An example of a PCIe device with single root input / output virtualization (SR-IOV) capability according to an embodiment of the present disclosure is shown;

[0016] Figure 6 A PCIe device according to an embodiment of the present disclosure is shown;

[0017] Figure 7 The physical functionality according to embodiments of this disclosure is illustrated;

[0018] Figure 8A A PCIe device in a link-up state according to an embodiment of the present disclosure is shown;

[0019] Figure 8B This illustrates setting the operating mode of a physical function to an active mode after the link is established, according to an embodiment of the present disclosure.

[0020] Figure 8C This illustrates setting the operating mode of a physical function to an inactive mode under normal operating conditions, according to an embodiment of the present disclosure.

[0021] Figure 8D This illustrates setting the operating mode of a physical function to an active mode under normal operating conditions, according to an embodiment of the present disclosure.

[0022] Figure 9A PCIe device according to another embodiment of the present disclosure is shown;

[0023] Figure 10 The process of operating a PCIe device according to an embodiment of the present disclosure is illustrated;

[0024] Figure 11 The process of providing PCIe device completion information according to embodiments of the present disclosure is illustrated.

[0025] Figure 12 The present disclosure illustrates a process for controlling a physical function that is set to an active mode, according to embodiments thereof; and

[0026] Figure 13 A process for controlling a physical function that has been set to an inactive mode, according to an embodiment of the present disclosure, is illustrated. Detailed Implementation

[0027] The description of the specific structure or function of examples of embodiments of the concept disclosed in this specification is merely for illustrating examples of embodiments of the concept, and examples of embodiments of the concept can be implemented in various forms, but these descriptions are not limited to the examples of embodiments described in this specification.

[0028] Figure 1 An example of a computing system 100 according to an embodiment of the present disclosure is shown.

[0029] Reference Figure 1 The computing system 100 may include a host 1000 and a high-speed peripheral component interconnect (PCIe) device 2000. The computing system 100 may be, for example, a cellular phone, a smartphone, an MP3 player, a laptop computer, a desktop computer, a game console, a television, a tablet PC, or an in-vehicle infotainment system.

[0030] The host 1000 may include a processor 1100, a memory 1200, an input / output (I / O) manager 1300, and a bus 1400. The components of the host 1000 can exchange signals and data through the bus 1400.

[0031] Processor 1100 may include circuitry, interfaces, or program code for processing data and controlling the operation of components of computing system 100. For example, processor 1100 may include a CPU (such as an ARM processor). TM Processors, microcontrollers, or application-specific integrated circuits (ASICs).

[0032] Memory 1200 may include SRAM or DRAM, which stores data, commands, or program code required for the operation of computing system 100. However, processor 1100 may include non-volatile memory. According to an embodiment, memory 1200 may include program code operable to run one or more operating systems (OS) and virtual machines (VMs), as well as program code running a virtualization intermediary (VI) for managing virtual machines.

[0033] Processor 1100 can run one or more operating systems and virtual machines via program code stored in drive memory 1200. Additionally, processor 1100 can run a virtualization intermediary for managing virtual machines. In this way, processor 1100 can control the operation of components of computing system 100.

[0034] I / O manager 1300 may be an adapter that connects input / output devices to host 1000. For example, I / O manager 1300 may include: a Universal Serial Bus (USB) adapter, a Peripheral Component Interconnect (PCI) or High Speed ​​PCI (PCIe) adapter, a Small Computer System Interface (SCSI) adapter, a Serial AT Attach (SATA) adapter, a High Speed ​​Non-Volatile Memory (NVMe) adapter, or a combination thereof. I / O manager 1300 may include circuitry, interfaces, or code operable to communicate with devices connected to computing system 100. I / O manager 1300 may include one or more standardized buses and one or more bus controllers. Therefore, I / O manager 1300 can sense devices connected to bus 1400, list devices connected to bus 1400, and perform resource allocation and deallocation for each device connected to bus 1400. In other words, I / O manager 1300 can operate to manage communication via bus 1400. For example, the I / O manager 1300 can be a PCI or PCIe system and includes a PCIe root union (RC), one or more PCIe switches or bridges. For example, the I / O manager 1300 can be controlled by a virtualization intermediary.

[0035] PCI can define a bus protocol for connecting I / O devices to the processor 1100. PCIe can have the programming concept defined by the PCI standard and define the physical communication layer as a high-speed serial interface.

[0036] PCIe device 2000 can communicate with host 1000 using PCIe. For example, PCIe device 2000 can be implemented as various I / O device types such as network and storage devices.

[0037] According to an embodiment, the PCIe device 2000 may be defined as an endpoint or a device including an endpoint.

[0038] An endpoint can refer to a function of the following types: it can be a requester or a completer of a PCIe transaction. Endpoints can be classified as traditional endpoints, high-speed PCI (PCIe) endpoints, or Root Consortium Integration (RCiEP) endpoints.

[0039] Traditional endpoints can be functional with a type 00h configuration space header. Traditional endpoints can act as completers to support configuration requests. Traditional endpoints can act as completers to support I / O requests. Traditional endpoints can accept I / O requests to either or both of the 80h and 84h locations, regardless of the corresponding endpoint's I / O decoding configuration. Traditional endpoints can support I / O requests. Traditional endpoints should not issue lock requests. Traditional endpoints can implement extended configuration space capabilities. Traditional endpoints operating as requesters of memory transactions may not need to generate 4GB or more of addresses. When requesting interruptible resources, traditional endpoints may need to support Message Signaled Interrupt (MSI) or MSI-X or both. When implementing MSI, traditional endpoints can support 32-bit or 64-bit message address versions with MSI functionalities. For the base address register of a requested memory resource, traditional endpoints can support 32-bit address specification. Traditional endpoints can appear within one of multiple hierarchical domains originating from the root union.

[0040] PCIe endpoints can be functional with a type 00h configuration space header. PCIe endpoints can act as completers to support configuration requests. PCIe endpoints should not rely on operating system allocations of I / O resources claimed via the base address register (BAR). PCIe endpoints may not generate I / O requests. PCIe endpoints cannot act as completers to support lock requests or as requesters to generate lock requests. PCIe-compliant software drivers and applications can be written to prevent the use of lock semantics when accessing PCIe endpoints. PCIe endpoints operating as requesters of memory transactions can generate addresses exceeding 4GB. When requesting interrupt resources, PCIe endpoints may need to support Message Signaled Interrupt (MSI), MSI-X, or both. When implementing MSI, PCIe endpoints can support a 64-bit message address version with MSI functionalities. The minimum address range requested by the base address register can be 128 bytes. PCIe endpoints can appear in a hierarchy domain originating from the root union.

[0041] RCiEP can be implemented on the internal logic of the root union containing the root port. RCiEP can be functional with a type 00h configuration space header. RCiEP can act as a completer to support configuration requests. RCiEP may not request I / O resources declared via the base address register. RCiEP may not generate I / O requests. RCiEP cannot act as a completer to support lock requests or generate lock requests as a requester. PCIe-compliant software drivers and applications can be written to prevent the use of locking semantics when accessing RCiEP. An RCiEP operating as a requester of a memory transaction can generate an address that is equal to or greater than an address that can be processed by the host 1000 as a completer. When requesting interrupt resources, RCiEP may be required to support MSI, MSI-X, or both. When implementing MSI, RCiEP can support 32-bit or 64-bit message address versions with MSI functionalities. For the base address register requesting memory resources, RCiEP can support 32-bit address specification. RCiEP can implement link capabilities, link states, link control, link capability 2, link state 2, and link control 2 registers within PCIe extension capabilities. RCiEP may not implement active state power management. RCiEP may not be hot-pluggable independently of the entire root federation. RCiEP may not appear in the hierarchical domain exposed by the root federation. RCiEP may not appear in switches.

[0042] According to an embodiment, the PCIe device 2000 can generate at least one virtual device. For example, the PCIe device 2000 can store program code for generating at least one virtual device.

[0043] According to an embodiment, the PCIe device 2000 can generate a physical function (PF) device or a virtual function (VF) device in response to a virtualization request received from the host 1000. For example, a physical function device can be configured as a virtual device for which a virtualization intermediary has been granted access. A virtual function device can be configured as a virtual device assigned to a virtual machine on the host 1000.

[0044] Figure 2 Another example of a computing system (computing system 200) according to an embodiment of the present disclosure is shown.

[0045] According to an embodiment, Figure 2 An example of a common platform configuration available for PCIe is shown.

[0046] Reference Figure 2The computing system 200 may include: a processor 1100, a memory 1200, multiple system images 210-1 to 210-n, a virtualization intermediary 220, a root union 230, a switch 240, and multiple PCIe devices 2000-1 to 2000-3.

[0047] Figure 2 The processor 1100 and memory 1200 shown can respectively represent Figure 1 The processor 1100 and memory 1200 are shown. According to an embodiment, the processor 1100, memory 1200, multiple system images 210-1 to 210-n, virtualization intermediary 220, root federation 230, and switch 240 may be included. Figure 1 In the host 1000. According to an embodiment, multiple PCIe devices 2000-1 to 2000-3 can represent Figure 1 PCIe devices 2000.

[0048] System images 210-1 to 210-n can refer to software components running in a virtual system allocated with PCIe functionality. According to an embodiment, system images 210-1 to 210-n can be referred to as virtual machines. System images 210-1 to 210-n can be software used to run applications or reliable services, such as an operating system. For example, each of system images 210-1 to 210-n can be configured as a guest OS, a shared or non-shared I / O device driver, etc. To increase effective hardware resource utilization without hardware modification, multiple system images 210-1 to 210-n can be run in the computing system 200.

[0049] According to embodiments, a PCIe function can refer to a separate operating unit used to provide physical resources included in PCIe devices 2000-1 to 2000-3. In this disclosure, PCIe function and "function" can have the same meaning.

[0050] Virtualization intermediary 220 may be a software component supporting one or more system images 210-1 to 210-n. According to embodiments, virtualization intermediary 220 may be referred to as a hypervisor or virtual machine monitor (VMM). Virtualization intermediary 220 may be interposed between hardware such as processor 1100 and memory 1200 and system images 210-1 to 210-n. I / O operations (inbound or outbound) in computing system 200 may be intercepted and processed by virtualization intermediary 220. Virtualization intermediary 220 may utilize its own virtual system to present abstract hardware resources to system images 210-1 to 210-n. The actual hardware resources available for each of system images 210-1 to 210-n may vary depending on workload or customer-specific policies.

[0051] Root union 230 can represent the root of an I / O hierarchy that connects the processor 1100 / memory 1200 subsystem to I / O. According to an embodiment, root union 230 can be implemented as follows: Figure 1 The diagram shows one configuration of the I / O manager 1300.

[0052] The computing system 200 may include at least one root union 230. Additionally, each root union 230 may include one or more root ports 231-1 and 232-2. Each of the root ports 231-1 and 232-2 may represent a separate hierarchical structure. The root union 230 may communicate with the switch 240 or PCIe devices 2000-1 to 2000-3 via the root ports 231-1 and 232-2.

[0053] The ability to route peer-to-peer transactions across multiple hierarchical domains via root union 230 is optional. A hierarchical domain can consist of a single endpoint or a sub-hierarchy containing one or more switches and endpoints.

[0054] When routing peer transactions between hierarchical domains, root union 230 may split packets into smaller packets. For example, root union 230 may split a single packet with a 256-byte payload into two packets, each with a 128-byte payload. However, root union 230 supporting peer routing of vendor-defined messages may not split vendor-defined messages into smaller packets, except at 128-byte boundaries (i.e., all resulting packets except the last one must be an integer multiple of 128 bytes).

[0055] Root union 230 can act as a requester to support the generation of configuration requests. Root union 230 can also act as a requester to support the generation of I / O requests.

[0056] Root union 230 cannot act as a completer to support locking semantics. Root union 230 can act as a requester to support the generation of lock requests.

[0057] Switch 240 can be defined as a logical component of each virtual PCI-to-PCI bridge. Switch 240 can communicate with PCIe devices 2000-2 and 2000-3.

[0058] Switch 240 can be provided as configuration software with two or more logical PCI-to-PCI bridges.

[0059] Switch 240 can use the PCI bridge mechanism to forward transactions. Switch 240 can forward all types of Transaction Layer Packets (TLPs) between all port sets. Switch 240 can support locking requests.

[0060] Switch 240 may not allow packets to be split into smaller packets.

[0061] When contention occurs on the same virtual channel, round-robin or weighted round-robin can be used to arbitrate between the ingress ports of switch 240.

[0062] The endpoint may not appear as a peer of a virtual PCI-to-PCI bridge representing a downstream port of the switch in the configuration software on the internal bus of the switch 240.

[0063] Figure 3 Another example of a computing system (computing system 300) according to an embodiment of the present disclosure is shown.

[0064] According to an embodiment, Figure 3 An example of a Single Root Input / Output Virtualization (SR-IOV) configuration platform is shown.

[0065] Single Root I / O Virtualization (SR-IOV) enables system images 210-1 to 210-n to share PCIe hardware resources in a virtual environment. According to an embodiment, SR-IOV can be developed to enhance the I / O capabilities of PCIe devices 2000-1 to 2000-3 and directly connect system images 210-1 to 210-n to PCIe devices 2000-1 to 2000-3. Therefore, in a computing system 300 incorporating SR-IOV capabilities, one or more PCIe devices 2000-1 to 2000-3 or PCIe functions can be assigned to one of the system images 210-1 to 210-n. According to an embodiment, SR-IOV capabilities can be specified such that a single PCIe device (2000-1 to 2000-3) can appear as a separate PCIe device 2000-1 to 2000-3 on a system image (210-1 to 210-n) under a single root port. According to the embodiment, PCIe devices 2000-1 to 2000-3 supporting SR-IOV capability can display multiple instances on system images 210-1 to 210-n, i.e., PCIe functions. Furthermore, in the computing system 300 employing SR-IOV capability, the root federation 230 can directly connect system images 210-1 to 210-n to PCIe functions without the virtualization intermediary 220. Therefore, by using SR-IOV capability, the root federation 230 can directly connect the system images 210-1 to 210-n of the host 1000 to PCIe functions without going through the virtualization intermediary 220.

[0066] Reference Figure 3 The computing system 300 may include: a processor 1100, a memory 1200, multiple system images 210-1 to 210-n, a virtualization intermediary 220, a root federation 230, a switch 240, a single root PCI manager (SR-PCIM) 310, a translation agent (TA) 320, an address translation and protection table (ATPT) 330, and multiple PCIe devices 2000-1 to 2000-3.

[0067] Compared to Figure 2 The computing system 200, Figure 3 The computing system 300 shown may further include: a single PCI manager 310, a translation agent 320, and an address translation and protection table 330.

[0068] like Figure 3 The multiple PCI devices shown, 2000-1 to 2000-3, can represent Figure 2Multiple PCIe devices 2000-1 to 2000-3. Additionally, each of the PCIe devices 2000-1 to 2000-3 may include: an address translation cache (ATC) 341, a physical function (PF0) 342, and multiple virtual functions (VF0_1 to VF0_N) 343. For ease of explanation, Figure 3 One physical function is shown. However, according to the embodiment, multiple physical functions may exist.

[0069] The single PCI manager 310 may include software responsible for configuring SR-IOV capabilities, managing physical functions 342 and virtual functions 343, handling associated error events, and overall device control such as power management and hot-plug services. According to an embodiment, the single PCI manager 310 may be logically located within the virtualization intermediary 220.

[0070] Translation agent 320 can be hardware or a combination of hardware and software responsible for translating addresses within a PCIe transaction into associated platform physical addresses. Translation agent 320 may include an address translation cache to accelerate translation table access. Additionally, translation agent 320 may support Address Translation Service (ATS), which enables PCIe functions to obtain address translations before DMA accesses the associated memory. In SR-IOV implementations, the use of translation agent 320 may be optional.

[0071] Address translation and protection table 330 may contain a set of address translations accessed by translation agent 320 to handle PCIe requests (DMA read, DMA write, or interrupt requests). In PCIe, interrupts can be treated as memory write operations. Interrupts can be routed to any target (e.g., processor core) that is transparent to the associated I / O function by a combination of the requester identifier and address contained within the PCIe transaction. DMA read and write requests can be translated by a combination of the route ID and the address contained within the PCIe transaction. The use of address translation and protection table 330 may be optional in SR-IOV implementations.

[0072] Address translation cache 341 can exist in two locations within the platform. For example, address translation cache 341 can be integrated within translation agent 320, located in RC, or located in PCIe device 2000-1 to 2000-3. In PCIe device 2000-1 to 2000-3, ATS technology can be used to input data to address translation cache 341. Instructions for PCIe transactions containing translated addresses can bypass the platform's address translation cache 341 to improve performance without compromising the advantages associated with address translation and protection table 330. The use of address translation cache 341 can be optional in SR-IOV implementations.

[0073] Physical function 342 may be a PCIe function that supports SR-IOV capability and accesses a single root PCI manager 310, virtualization intermediary 220, or system images 210-1 to 210-n.

[0074] Virtual function 343 can be a lightweight PCIe function that can directly access system images 210-1 to 210-n. Virtual function 343 can operate as a virtual instance of physical function 342. Therefore, from the perspective of system images 210-1 to 210-n, virtual function 343 can appear as an instance of physical function 342. Resources associated with the main data movement of the function can be used in system images 210-1 to 210-n. Virtual function 343 can be shared sequentially by different system images 210-1 to 210-n. For example, virtual function 343 can be assigned to one system image (210-1 to 210-n), can be reset, and can then be assigned to another system image (210-1 to 210-n). Virtual function 343 can be selectively migrated from one physical function 342 to another physical function. All virtual functions 343 associated with physical function 342 may have the same device type as physical function 342 (e.g., the same network device type or the same storage device type).

[0075] Figure 4 A multifunctional device according to an embodiment of the present disclosure is shown.

[0076] Figure 4 The PCIe device 400 shown can represent one of the following: Figure 1 The PCIe device 2000 shown and Figure 2 and Figure 3 The PCIe devices shown are 2000-1 to 2000-3.

[0077] Reference Figure 4 The PCIe device 400 may include multiple functions, Function 0 to Function N. According to an embodiment, the multiple functions, Function 0 to Function N, may be physical functions. When all functions, Function 0 to Function N, included in the PCIe device 400 are physical functions, the corresponding PCIe device 400 may be defined as a multi-functional device.

[0078] According to an embodiment, multiple functions, from Function 0 to Function N, can share a common PCIe link. The link and PCIe functionality shared by all functions, from Function 0 to Function N, can be managed through Function 0 (the zeroth function).

[0079] According to an embodiment, multiple functions, from Function 0 to Function N, can support Alternative Routing Identifier (ARI) capabilities. Additionally, multiple ARI capabilities can utilize a single bus number captured via a PCI enumeration process.

[0080] According to an embodiment, multiple functions 0 to N can support ATS capabilities. These functions 0 to N may include an address translation cache for managing translated addresses obtained by the ATS.

[0081] According to an embodiment, each of the multiple functions 0 to N may include unique physical resources, such as a separate configuration space, a base address register (BAR), etc. The physical resources may refer to some of the physical resources of the PCIe device 400. In other words, the multiple functions 0 to N can provide physical resources to the allocated system image.

[0082] Multiple functions, from Function 0 to Function N, can be assigned to system images separately. To prevent one system image from affecting another, all PCIe configuration operations can be intercepted and processed by the virtualization intermediary.

[0083] According to an embodiment, the configuration space of each of the plurality of functions 0 to N may include a type 0 configuration space header. For example, the type 0 configuration space header may include a header type register and an interrupt pin register.

[0084] A header type register can be included in the type 0 configuration space header of the configuration space. The header type register can identify whether multiple functions can be included in the predefined second part (starting from byte 10h of the configuration space) and the layout of the PCIe device 400. For example, a multifunction device field can be included in the header type register. When the multifunction device field is set, the PCIe device 400 can include multiple functions, Function 0 through Function N. Software can detect functions 1 through N other than function 0.

[0085] The interrupt pin register can be a read-only register that identifies the legacy interrupt message used by the function. Valid values ​​can be 01h, 02h, 03h, and 04h, which map to the legacy interrupt messages INTA, INTB, INTC, and INTD, respectively. The value 00h indicates that the function does not use a legacy interrupt message. Values ​​05h to FFh can be reserved. High-speed PCI can define one legacy interrupt message for a single-function device and up to four legacy interrupt messages for a multi-function device. For a single-function device, only INTA can be used. All functions (Function 0 to Function N) of a multi-function device can use the INTx message. When a device implements a single legacy interrupt message, it should be INTA. When two legacy interrupt messages are implemented, they should be INTA and INTB. For a multi-function device, all functions (Function 0 to Function N) can use the same INTx, or each can have its own INTx (up to four), or a combination thereof.

[0086] According to the embodiment, the configuration space for all functions 0 to N can be included in the PCIe capability structure. For example, the PCIe capability structure may include: PCIe capability register, device capability register, device control register, device status register, link capability register, link control register, link status register, slot capability register, slot control register, slot status register, root control register, root capability register, root status register, device capability 2 register, device control 2 register, device status 2 register, link capability 2 register, link control 2 register, link status 2 register, slot capability 2 register, slot control 2 register, slot status 2 register, root controller 2 register, root capability 2 register, root status 2 register, etc.

[0087] The PCIe capability register identifies the function type and associated capabilities of a PCIe device. According to an embodiment, the PCIe capability register can be a 16-bit register. Bits 3:0 of ​​the PCIe capability register can be capability version fields. The capability version field can be read-only. More specifically, the capability version field can indicate the version number of the high-speed PCI function architecture defined by PCI-SIG. Bits 7:4 of the PCIe capability register can be device / port type fields. The device / port type field can be read-only. More specifically, the device / port type field can indicate a specific type of function. In the device / port type field, the function of a multi-function device can refer to different types. Bit 8 of the PCIe capability register can be a slot implementation field. The slot implementation field can have a hardware initialization (HwInit) attribute. When the slot implementation field is set, it can indicate that a link connected to that port is connected to the slot. Bits 13:9 of the PCIe capability register can be interrupt message number fields. The interrupt message number field can be read-only. The Interrupt Message Number field indicates which MSI / MSI-X vector is used for the interrupt message generated in association with a status bit having a functional structure. For MSI, the Interrupt Message Quantity field indicates the offset between the base message data and the generated interrupt message. For MSI-X, the Interrupt Message Number field indicates the MSI-X table entry used to generate the interrupt message. When both MSI and MSI-X are implemented, although only one mechanism can be enabled by the software at a time, different vectors can be used for MSI and MSI-X. For example, when MSI-X is enabled, the Interrupt Message Number field indicates the MSI-X vector. When MSI is enabled or neither MSI nor MSI-X is enabled, the Interrupt Message Number field indicates the MSI vector. When both MSI and MSI-X are enabled, the value in the Interrupt Message Number field may be undefined. Bit 14 of the PCIe Capability Register may have a read-only attribute and may be undefined.

[0088] The device capability register identifies function-specific capabilities of a PCIe device. According to an embodiment, the device capability register may be a 32-bit register. Bits 2:0 of ​​the device capability register may be a Max_Payload_Size Supported field. The Max_Payload_Size Supported field may have a read-only attribute. The Max_Payload_Size Supported field may indicate the maximum payload size supported for TLP. For this field, Function 0 through Function N of a multifunction device may have different values. Bits 4:3 of the device capability register may be a phantom function supported field. The phantom function supported field may have a read-only attribute. The phantom function supported field may indicate support for the use of undeclared function numbers to extend the number of allowed incomplete transactions by logically combining the undeclared function numbers with tag identifiers.

[0089] The device control register controls PCIe device-specific parameters. According to an embodiment, the device control register can be a 16-bit register. Bit 0 of the device control register can be a correctable error reporting enable field. The correctable error reporting enable field can be combined with other bits to control the transmission of ERR_COR messages. For a multifunction device, the correctable error reporting enable field can control error reporting for each function (Function 0 to Function N). The correctable error reporting enable field can have a read-write (RW) attribute. Bit 1 of the device control register can be a non-fatal correctable error reporting enable field. The non-fatal correctable error reporting enable field can be combined with other bits to control the transmission of ERR_NONFATAL messages. For a multifunction device, from the perspective of each function (Function 0 to Function N), the non-fatal error reporting enable field can control error reporting for each function (Function 0 to Function N). The non-fatal error reporting enable field can have a read-write attribute. Bit 2 of the device control register can be a serious correctable error reporting enable field. The Critical Correctable Error Reporting Enable field can be combined with other bits to control the sending of ERR_FATAL messages. For multifunction devices, from the perspective of each function (Function 0 to Function N), the Critical Correctable Error Reporting Enable field controls error reporting for each function (Function 0 to Function N). The Critical Correctable Error Reporting Enable field can have read / write attributes. Bit 3 of the device control register can be the Unsupported Request Reporting Enable field. The Unsupported Request Reporting Enable field can be combined with other bits to control the signaling of unsupported request errors by sending error messages. For multifunction devices, from the perspective of each function (Function 0 to Function N), the Unsupported Request Reporting Enable field controls error reporting for each function (Function 0 to Function N). The Unsupported Request Reporting Enable field can have read / write attributes. Bits 7:5 of the device control register can be the Max_Payload_Size field. The Max_Payload_Size field can set the maximum TLP payload size for the function. As a receiver, the function cannot generate a TLP exceeding the set value. As a transmitter, the allowed value can be indicated by the Maximum Payload Size Support field of the Function Capability Register. All functions of the multifunction device, Function 0 to Function N, can have different values ​​in the Maximum Payload Size field.For the Replacement Route Identifier (ARI) device, the Maximum Payload Size field can be determined solely by the settings in Function 0. Settings in other functions can always return any value software-programmed for each function, or otherwise be ignored by the component. The Maximum Payload Size field can have read and write attributes.

[0090] The device status register can provide information about PCIe device (function) specific parameters. According to an embodiment, the device status register can be a 16-bit register. Bit 0 of the device status register can be a correctable error detection field. The correctable error detection field can indicate that a correctable error has been detected. Correctable errors can be logged in the correctable error detection field regardless of whether error reporting is enabled in the device control register. For a multifunction device, each function (Function 0 to Function N) can indicate the error state perceived by each function (Function 0 to Function N). The correctable error detection field can have a write-1 to clear (RW1C) status attribute. Bit 1 of the device status register can be a non-critical correctable error detection field. The non-critical correctable error detection field can indicate that a non-critical error has been detected. Non-critical errors can be logged in the non-critical error detection field regardless of whether error reporting is enabled in the device control register. For a multifunction device, each function (Function 0 to Function N) can indicate the error state perceived by each function (Function 0 to Function N). The non-critical error detection field can have the RW1C attribute. Bit 2 of the device status register can be a critical error detection field. The critical error detection field can indicate the state in which a critical error has been detected. A critical error can be logged in the critical error detection field regardless of whether error reporting is enabled in the device control register. For a multifunction device, each function (Function 0 to Function N) can indicate the error state perceived by each function (Function 0 to Function N). The critical error detection field can have the RW1C attribute. Bit 3 of the device status register can be an unsupported request field. The unsupported request detection field can indicate that a function has received an unsupported request. An error can be logged in the unsupported request detection field regardless of whether error reporting is enabled in the device control register. For multi-function devices, each function (Function 0 to Function N) can indicate an error state perceived by each function (Function 0 to Function N). The Unsupported Request Detection field can have the RW1C attribute. Bit 6 of the device status register is the Emergency Power Saving Detection field, which can be set when Function 0 to Function N enters an emergency power saving state. Functions 0 to Function N can maintain the emergency power saving state whenever the conditions for entering it exist.The multifunction device connected to the upstream port can set an emergency power saving detection field in all functions (Function 0 to Function N) that support emergency power saving mode. The emergency power saving detection field can have the RW1C attribute.

[0091] The link capability register identifies the specific capabilities of a PCIe link. For multifunction devices, the field values ​​of the link capability registers for all functions (Function 0 through Function N) can be the same.

[0092] According to an embodiment, the configuration space for each of the multiple functions 0 to N may include: Message Signal Interruption (MSI) capability structure, assisted high-speed PCI extension capability, data link feature extension capability, and ACS extension capability. All interrupt-generating functions 0 to N can implement MSI or MSI-X.

[0093] In addition to the registers mentioned above, the configuration space for each of the multiple functions, from Function 0 to Function N, may also include individual registers associated with the multifunction device.

[0094] According to an embodiment, multiple functions 0 to N can perform error handling operations. For example, multiple functions 0 to N can use Advanced Error Reporting (AER) to perform error handling operations. For example, when an error is detected, multiple functions 0 to N can perform error handling operations by: indicating the error status through a completion status field, sending an error message to the root union, or forwarding the error.

[0095] According to an embodiment, multiple functions 0 to N can perform a function-level reset (FLR) operation. For example, multiple functions 0 to N can perform a function-level reset (FLR) operation upon request from the host.

[0096] Function-level reset mechanisms enable software to quiesce and reset endpoint hardware at a function-level granularity.

[0097] Function-level resets can be applied to each of Functions 0 through N. Only the target function (Function 0 through Function N) can be affected by a function-level reset operation. Link state is unaffected by function-level resets.

[0098] Function-level reset can modify the state of Function 0 to Function N.

[0099] For example, in addition to sticky type registers (ROS, RWS, RW1CS), HwInit type definition registers, and other such fields or registers, function registers and function-specific state machines can be set to their initial values.

[0100] Additionally, a function-level reset can modify the state of functions such as: the capture slot power limit value of the device capability register, the capture slot power limit scale of the device capability register, the Max_Payload_Size of the device control register, the active state power management (ASPM) control of the link control register, the read completion boundary (RCB) of the link control register, the general clock configuration of the link control register, the extended synchronization of the link control register, the enabled clock power management of the link control register, the hardware autonomous width disable of the link control register, the hardware autonomous speed disable of the link control 2 register, the link equalization 8.0GT / s request of the link state 2 register, the link equalization request 16.0GT / s of the 16.0GT / s state register, the channel equalization control register of the auxiliary high-speed PCI expansion capability structure, the 16.0GT / s channel equalization control register of the physical layer 16.0GT / s expansion capability structure, the virtual channel capability structure of all registers, all registers of the multi-function virtual channel capability structure, all registers of the data link feature expansion capability structure, all registers of the physical layer 16.0GT / s expansion capability structure, and all registers of the channel margin of the receiver expansion capability structure.

[0101] It can clear the control that enables functions 0 to N to initiate requests for high-speed PCI, including bus master enable, MSI interrupt enable, etc., thereby effectively making functions 0 to N stand still on the link.

[0102] The port state machine associated with a link function (including port state machines in the physical and data link layers) may not be reset by the FLR, and VC0 may remain initialized after the FLR. Any uncompleted INTx interrupt asserted by Function 0 through Function N can be deasserted by sending the corresponding Deassert_INTx message before the FLR begins.

[0103] When the FLR is initialized to Function 0 through Function N of PCIe device 400, the Deassert_INTx message will not be transmitted if another function continues to make the matching INTx effective.

[0104] After initializing the FLR by writing 1b to the initialization function-level reset bit, functions 0 through N must complete the FLR within 100ms. If the software initializes the FLR with the transaction pause bit set to 1b, it cannot initialize functions 0 through N until sufficient time is allowed for any associated completions or reasonable determinism is achieved (i.e., no remaining completions will occur). Therefore, the computing system can allow as much time as the pre-FLR value set for the completion timeout. If completion timeouts are disabled on functions 0 through N when publishing the FLR, the delay can be system-dependent but can be 100ms or longer. When implementing the function-ready state, the computing system can publish configuration requests to functions 0 through N immediately after receiving a function-level reset message indicating that functions 0 through N are fully ready for configuration.

[0105] Upon receiving an FLR, Functions 0 through N can clear all transaction states (including transaction pauses), or set the completion timeout to its default value so that all paused transactions time out during the FLR's execution. When the FLR completes, the transaction pause bit can be cleared.

[0106] Because the FLR modifies the function state, the behavior of the FLR can be characterized using the following set of criteria: when applied to functions 0 through N, it indicates that functions 0 through N have a modified state. The following criteria must be applied using function-specific knowledge to evaluate the behavior of functions 0 through N in response to the FLR.

[0107] - Functions (Function 0 through Function N) may not expose the adapter initialized by the active host on any external interface controlled by the corresponding function (Function 0 through Function N). Exceptions may be made to the steps required to terminate activity on the external interface. For example, for queries requiring adapter initialization by the host system or interaction with the active host system, the network adapter may not respond, but the network adapter may be allowed to perform actions designed to be performed without host initialization or interaction. When a network adapter includes multiple functions (Function 0 through Function N) running on the same external network interface, this rule may only affect those aspects associated with the specific function reset by the FLR.

[0108] Functions 0 through N may not maintain a software-readable state within themselves, which may include confidential information associated with any previous use of Functions 0 through N. The main host memory allocated to Functions 0 through N may not be modified by Functions 0 through N. For example, functions (Functions 0 through N) with internal memory that can be directly or indirectly read by host software may clear or randomize that memory.

[0109] - Functions (Function 0 to Function N) can return to a state that allows the normal configuration of the high-speed PCI interface of functions (Function 0 to Function N) to make it available to the drivers typically associated with functions (Function 0 to Function N).

[0110] When the FLR is initialized, target functions 0 through N must be performed as follows:

[0111] Functions 0 through N can return completion of the configuration writes that initialize the FLR operation, and then the FLR can be initialized.

[0112] - When an FLR is in progress, if a request arrives, it is permissible to silently discard the request without logging or signaling an error. Additionally, if completion arrives, it is permissible to treat the completion as an unexpected completion or silently discard it without logging or signaling an error.

[0113] When functions (Function 0 through Function N) are required to complete an FLR operation within a time limit, subsequent function-specific initialization sequences may require additional time. When additional time is required, functions (Function 0 through Function N) may return to the Configuration Request Retry Status (CRS) completion status when a configuration request is received after that time limit. After functions (Function 0 through Function N) respond to a configuration request with a completion status other than CRS, returning to CRS may not be allowed until it is reset again.

[0114] Figure 5 An example of a PCIe device 500 with single root input / output virtualization (SR-IOV) capability according to an embodiment of the present disclosure is shown.

[0115] Figure 5 The PCIe device 500 shown can represent one of the following: Figure 1 The PCIe device 2000 shown and Figure 2 and Figure 3 The PCIe devices shown are 2000-1 to 2000-3.

[0116] For example, Figure 5 This can be represented as a PCIe device 500 with SR-IOV capability.

[0117] Reference Figure 5 The PCIe device 500 may include a physical function PF0 and a plurality of virtual functions VF0_1 to VF0_N associated with the physical function PF0. Virtual function VF0_1 may refer to a first virtual function associated with the physical function PF0. Similarly, virtual function VF0_N may refer to an Nth virtual function associated with the physical function PF0. According to an embodiment, the above references... Figure 4 The described configuration space, register setting processing, error handling operations, and function-level reset operations can be applied to physical function PF0.

[0118] Physical function PF0 can support SR-IOV capability. According to an embodiment, physical function PF0 can disable SR-IOV capability after a reset operation. To discover the page sizes supported by physical function PF0 and the multiple virtual functions VF0_1 to VF0_N, the supported page size configuration field can be read.

[0119] Multiple virtual functions VF0_1 through VF0_N can share multiple common configuration space fields with physical function PF0. Sharing reduces the hardware resource requirements for implementing multiple virtual functions VF0_1 through VF0_N. For example, multiple virtual functions VF0_1 through VF0_N can use the same configuration mechanism and header type as physical function PF0. Additionally, multiple virtual functions VF0_1 through VF0_N can share the VF BAR set. Furthermore, multiple virtual functions VF0_1 through VF0_N can share the VF Storage Space Enable (MSE) bit, which controls access to the VF storage space. In other words, when the VF MSE bit is cleared, the storage mapping space allocated to all virtual functions VF0_1 through VF0_N can be disabled.

[0120] The InitialVFs and TotalVFs fields included in the SR-IOV capability can be used to discover the maximum number of virtual functions that can be associated with physical function PF0. For example, when PCIe device 500 does not support VF migration, TotalVFs and InitialVFs can contain the same value. On the other hand, when PCIe device 500 supports VF migration, if TotalVFs is read, physical function PF0 can return the number of virtual functions that can be allocated to physical function PF0. Additionally, when InitialVFs is read, physical function PF0 can return the initial number of virtual functions allocated to physical function PF0.

[0121] Physical function PF0 and multiple virtual functions VF0_1 to VF0_N can be assigned unique route IDs. The route ID for each virtual function (VF0_1 to VF0_N) can be determined using the route ID of physical function PF0 and fields included in the SR-IOV capability of physical function PF0.

[0122] All PCIe and SR-IOV configuration access can be performed through trusted software components such as virtualization intermediaries or single root PCI managers.

[0123] According to an embodiment, physical function PF0 and multiple virtual functions VF0_1 to VF0_N may contain physical resources including configuration space. Each virtual function (VF0_1 to VF0_N) may contain a set of non-shared physical resources required to deliver function-specific services, such as resources like work queues and data buffers. These resources can be directly accessed by the system image without the intervention of a virtualization intermediary or a single root PCI manager.

[0124] One or more virtual functions VF0_1 to VF0_N can be assigned to each system image. The physical function PF0 and multiple virtual functions VF0_1 to VF0_N may include ATC.

[0125] According to an embodiment, the physical function PF0 can be used to manage device-specific functions, such as internal resource allocation for each virtual function (VF0_1 to VF0_N) and VF arbitration of shared resources for such PCIe links or function-specific links (e.g., network or storage links).

[0126] According to an embodiment, the configuration space of physical function PF0 may include SR-IOV extension capabilities. For example, SR-IOV extension capabilities may include: SR-IOV extension capability header register, SR-IOV capability register, SR-IOV control register, SR-IOV status register, InitialVFs register, TotalVFs register, NumVFs register, function-dependent link register, first VF offset register, VF step size register, VF device ID register, supported page size register, system page size register, VF BAR0, VF BAR1, ..., VF BAR5 registers, and VF migration state array offset register.

[0127] The SR-IVO control register may include a Virtual Function (VF) enable field.

[0128] Virtual function enabling allows management of the allocation of multiple virtual functions VF0_1 to VF0_N to their associated physical function PF0. When virtual function enabling is set, multiple virtual functions VF0_1 to VF0_N associated with physical function PF0 can be accessed within the High-Speed ​​PCI architecture. When virtual function enabling is set, virtual functions VF0_1 to VF0_N can respond to and publish High-Speed ​​PCI transactions in accordance with the rules of High-Speed ​​PCI endpoint functions.

[0129] When virtual functions are cleared, several virtual functions VF0_1 to VF0_N can be disabled and may not be visible in the High-Speed ​​PCI architecture. Requests for these virtual functions VF0_1 to VF0_N may receive Unsupported Requests (URs) and may not be published for High-Speed ​​PCI transactions.

[0130] To allow components to perform internal initialization, after changing the virtual function enable bit from 0 to 1, the system may not be allowed to issue requests to virtual functions VF0_1 through VF0_N enabled by the corresponding enable bit until one of the following is met: at least 100 milliseconds have elapsed; an FRS message with a reason code for virtual function enable has been received from physical function PF0; and at least the VF enable time has elapsed. The VF enable time can be a reset time value in the readiness time reporting capability associated with virtual functions VF0_1 through VF0_N, or a value determined by the system software / firmware.

[0131] Before the root consortium and / or system software can determine that a virtual function (VF0_1 to VF0_N) is corrupt and unable to return a successful completion status for a valid configuration request, the root consortium and / or system software may allow at least 1.0 seconds after setting the VF enable bit. After setting the VF enable bit, when the virtual functions VF0_1 to VF0_N enabled by the corresponding VF enable bit are not yet ready to provide a successful completion status for a valid configuration request, virtual functions VF0_1 to VF0_N may be allowed to return the CRS status to the configuration request for a maximum of 1.0 seconds. After physical function PF0 sends an FRS message with a reason code for VF enable, virtual functions VF0_1 to VF0_N associated with the corresponding physical function PF0 are not allowed to return CRS without intervention of VF disable or other valid reset conditions. After returning successful completion to any request, virtual functions VF0_1 to VF0_N may not be allowed to return CRS without intervention of VF disable or other valid reset conditions.

[0132] Because virtual functions VF0_1 to VF0_N do not have the MSE bit (the MSE in virtual functions VF0_1 to VF0_N is controlled by the VF MSE bit in the SR-IOV capability of physical function PF0), software can issue storage requests before the virtual functions (VF0_1 to VF0_N) are ready to process the MSE bit. Therefore, storage requests cannot be issued to virtual functions VF0_1 to VF0_N until at least one of the following conditions is met:

[0133] - The virtual functions (VF0_1 to VF0_N) have successfully responded to the configuration request (without returning a CRS).

[0134] - After publishing an FLR to a VF, at least one of the following conditions must be met: (1) at least 1.0 seconds have elapsed since the FLR was published; (2) virtual functions VF0_1 to VF0_N are in a ready state, and an FRS message with a reason code for FLR completion has been received from virtual functions VF0_1 to VF0_N after the FLR was published; or (3) at least the FLR time has elapsed since the FLR was published. The FLR time may be the FLR time value in the readiness time reporting capability associated with virtual functions VF0_1 to VF0_N, or a value determined by the system software / firmware.

[0135] - After VF enable is set in physical function PF0, at least one of the following is true: (1) at least 1.0 second has elapsed since VF enable was set; (2) physical function PF0 supports the function ready state, and after VF enable is set, an FRS message with a reason code for virtual function enable has been received from physical function PF0; or (3) after virtual function enable is set, the minimum virtual function enable time has elapsed. The virtual function enable time may be a reset time value in the ready time reporting capability associated with virtual functions VF0_1 to VF0_N, or a value determined by the system software / firmware.

[0136] After an FLR is issued to virtual functions VF0_1 through VF0_N or after VF is enabled in the SR-IOV capability in the associated physical function PF0, virtual functions VF0_1 through VF0_N can be allowed to silently drop storage requests until the virtual functions VF0_1 through VF0_N successfully respond to any request (without returning a CRS).

[0137] Clearing VF enable effectively destroys virtual functions VF0_1 to VF0_N. Setting VF enable effectively creates virtual functions VF0_1 to VF0_N. Setting the VF enable bit after VF enable has been previously cleared generates a new set of virtual functions VF0_1 to VF0_N. When physical function PF0 is in D0 power state, this new set of virtual functions VF0_1 to VF0_N can be in D0 power state. 未初始化 (D0 uninitialized ) State. If the physical function PF0 is in a low-power state, the behavior may not be defined.

[0138] When VF is cleared, a physical function PF0 that supports FRS can send an FRS message with the reason for VF disabling to indicate when the operation is complete. If there are incomplete non-posted requests issued by physical function PF0 or any virtual functions VF0_1 to VF0_N associated with physical function PF0, physical function PF0 may not be allowed to send this message. The FRS message can only be sent after these requests have completed (or timed out).

[0139] When software clears VF enable, fields in the SR-IOV extension capability or VF migration state array become inaccessible until:

[0140] - Wait at least 1.0 second after clearing VF enable, or

[0141] - The physical function PF0 supports FRS, and after clearing VF enable, an FRS message with a reason code for VF disable has been received from the physical function PF0.

[0142] The InitialVFs register can indicate to SR-PCIM the number of virtual functions VF0_1 to VF0_N initially associated with physical function PF0.

[0143] The InitialVFs register can indicate the maximum number of virtual functions VF0_1 to VF0_N that can be associated with physical function PF0.

[0144] The NumVFs register controls the number of visible virtual functions VF0_1 to VF0_N.

[0145] The Virtual Function Base Address Register (BAR) can define the base address registers (BARs) for virtual functions VF0_1 to VF0_N. When an actual address value is written to the Virtual Function Base Address Register and VF Enable and VF MSE are set, the BAR can map NumVFs BARs.

[0146] According to the embodiments, the configuration space of physical function PF0 and virtual functions VF0_1 to VF0_N may include a type 0 configuration space header, PCIe capability register, PCI standard capability register, PCIe extended capability register, etc.

[0147] According to an embodiment, the register field values ​​included in the configuration space of virtual functions VF0_1 to VF0_N can be either dependent on or independent of the register field values ​​included in the configuration space of physical function PF0. For example, when the register field values ​​included in the configuration space of virtual functions VF0_1 to VF0_N are dependent on the register field values ​​included in the configuration space of physical function PF0, the register field values ​​included in the configuration space of virtual functions VF0_1 to VF0_N can have a value of 0 (zero). Virtual functions VF0_1 to VF0_N can perform configuration operations based on the register field values ​​of the associated physical function PF0.

[0148] According to an embodiment, multiple virtual functions VF0_1 to VF0_N can perform error handling operations. For example, when physical function PF0 uses advanced error reporting to handle errors, virtual functions VF0_1 to VF0_N can also use advanced error reporting to handle errors.

[0149] According to an embodiment, multiple virtual functions VF0_1 to VF0_N can perform a reset operation.

[0150] For example, virtual functions VF0_1 to VF0_N can complete internal initialization when one of the following conditions occurs:

[0151] - The virtual functions (VF0_1 to VF0_N) have successfully responded to the configuration request (no CRS returned).

[0152] - After a function-level reset is issued to virtual functions VF0_1 through VF0_N, at least one of the following is true: (1) at least 1.0 second has elapsed; (2) an FRS message with a cause code for FLR completion has been received from virtual functions VF0_1 through VF0_N; or (3) the minimum FLR time has elapsed. The function-level reset time may be a function-level reset time value in the readiness time reporting capability associated with virtual functions VF0_1 through VF0_N, or a value determined by the system software / firmware.

[0153] - After VF is enabled in physical function PF0, at least one of the following can be true: (1) at least 1.0 seconds have passed since VF was enabled; or (2) after VF is enabled, an FRS message with a reason code for VF being enabled has been received from physical function PF0.

[0154] Regarding the above reference Figure 4 The described multi-functional device may not be able to control the number of physical functions, and may not hide the physical functions when the client OS needs to terminate the operation. Similarly, regarding the reference... Figure 5 The described virtual functions may require disabling all virtual functions to control their number. Furthermore, because a large portion of a virtual function's configuration space is influenced by its associated physical function, it may be impossible to perform PCIe configuration settings for each virtual function. Therefore, a PCIe device that combines the advantages of both physical and virtual functions is needed.

[0155] Figure 6 A PCIe device 600 according to an embodiment of the present disclosure is shown.

[0156] PCIe device 600 can communicate with host 1000 via PCIe link.

[0157] Reference Figure 6 The PCIe device 600 may include physical functions (PF)0 610, physical functions 1 620-1 to physical functions n 620-n, and a function mode controller 630. The PCIe device 600 can represent... Figure 1 The PCIe device shown is 2000 or may be Figure 2 and Figure 3 The PCIe device shown is one of 2000-1 to 2000-3. Physical function 0 610 can have the same features as the one referenced above. Figure 4 The physical function PF0 described and the above reference Figure 5The physical function PF0 described has the same configuration. In this specification, physical function 0 610 may be referred to as the "first physical function". In addition, in this specification, physical functions 1 620-1 to physical functions n 620-n may be referred to as "a plurality of second physical functions".

[0158] Physical functions 0 610 and 1 620-1 through 620-n can be PCIe functions. Each of physical functions 0 610 and 1 620-1 through 620-n can be assigned to one of a plurality of system images included in host 1000. The plurality of system images can access the assigned physical functions of physical functions 610 and 620-1 through 620-n.

[0159] According to an embodiment, each of physical function 0 610 and physical functions 1 620-1 to 620-n can be an independent operating unit for providing physical resources included in the PCIe device 600. Each of physical function 0 610 and physical functions 1 620-1 to 620-n can be allocated to some of the plurality of physical resources included in the host 1000. Each of physical function 0 610 and physical functions 1 620-1 to 620-n can provide corresponding physical resources of the PCIe device 600 and the host 1000 to each of the allocated system images.

[0160] Physical function 0610 can be a function that must be included in the PCIe device 600. According to an embodiment, physical function 0610 can be a function that cannot be disabled.

[0161] According to an embodiment, physical function 0 610 can manage the PCIe links connected to host 1000. The PCIe links can be general-purpose links shared by physical function 0 610 and physical functions 1 620-1 to 620-n. Each of physical function 0 610 and physical functions 1 620-1 to 620-n can transmit and receive data, data packets, etc., to and from the assigned system image via the PCIe links.

[0162] Additionally, physical function 0 610 can manage PCIe functions. PCIe functions can be common functions shared by physical function 0 610 and physical functions 1 620-1 to physical functions n 620-n.

[0163] According to an embodiment, physical function 0 610 can receive function mode control information MODE_INFO from host 1000. Physical function 0 610 can provide function mode control information MODE_INFO to function mode controller 630.

[0164] The function mode control information MODE_INFO may include information for setting the operating modes of physical functions 1620-1 to 620-n. According to an embodiment, the function mode control information MODE_INFO may include at least one of the following: the number of physical functions to be enabled among physical functions 1620-1 to 620-n, identification information of the physical functions to be enabled, the number of physical functions to be disabled among physical functions 1620-1 to 620-n, and identification information of the physical functions to be disabled. For example, when multiple physical functions are to be enabled, the function mode control information MODE_INFO may include the number of physical functions to be enabled among physical functions 1620-1 to 620-n and identification information of the physical functions to be enabled. In another example, when multiple physical functions are to be disabled, the function mode control information MODE_INFO may include the number of physical functions to be disabled among physical functions 1620-1 to 620-n and identification information of the physical functions to be disabled. In another example, when enabling a physical function, the function mode control information MODE_INFO may include identification information of the physical function to be enabled from physical function 1 620-1 to physical function n 620-n. In another example, when disabling a physical function, the function mode control information MODE_INFO may include identification information of the physical function to be disabled from physical function 1 620-1 to physical function n 620-n. According to an embodiment, the identification information of the physical function may include at least one of a bus number, a device number, and a function number.

[0165] According to the example, physical function 0 610 can use one of the following to receive function mode control information MODE_INFO from host 1000: vendor-defined message, vendor-defined capability, new PCIe capability defined by the user, Management Component Transport Protocol (MCTP) on System Manager Bus (SMbus), and MCTP on PCIe.

[0166] According to an embodiment, vendor-defined messages can allow for extended high-speed PCI messaging capabilities. For example, a vendor-defined message may include a requester ID field for sending a message, a receiver ID field for receiving a message, and vendor-defined fields. The receiver ID may include at least one of a bus number, a device number, and a function number. For example, host 1000 can provide function mode control information MODE_INFO to physical function 0 610 through the vendor-defined fields of the vendor-defined message. The requester ID may include identification information of host 1000 or the system image. The receiver ID may be identification information of the PCIe device 600 or the physical function receiving the message.

[0167] According to an embodiment, a vendor-specific capability can be a capability structure within a PCI-compatible configuration space. Vendor-specific capabilities allow device vendors to use capability mechanisms for vendor-specific information. A vendor-specific capability register may include a capability ID field, a next capability pointer field, a capability length field, and a vendor-specific information field. The capability ID field may indicate a PCIe capability structure. The capability ID field can provide a capability ID indicating a vendor-specific capability structure. The next capability pointer field may indicate the offset of the next PCI capability structure. When no linked list of capabilities exists, the next capability pointer field may include 00h. The capability length field can provide the number of bytes included in the capability structure. The vendor-specific information field can be changed depending on the vendor. Therefore, the vendor-specific information field can be set to various values ​​according to the user. For example, host 1000 can provide function mode control information MODE_INFO to physical function 0610 through the vendor-specific capability's specific information field.

[0168] According to an embodiment, a new PCIe capability defined by the user may include identification information of the device providing the information, identification information of the device receiving the information, and fields including the information.

[0169] According to an embodiment, physical functions 1 620-1 to 620-n can be enabled or disabled according to the operating mode. For example, physical functions 1 620-1 to 620-n can be set to one of an active mode or an inactive mode. Physical functions 1 620-1 to 620-n set to active mode can be enabled to perform operations and can perform communication with the allocated system image via a PCIe link. Physical functions 1 620-1 to 620-n set to inactive mode can be disabled so that no operations are performed and communication is not performed via a PCIe link.

[0170] According to an embodiment, the function mode controller 630 can set the operating mode of each of physical functions 1 620-1 to physical functions n 620-n to one of the active mode and inactive mode based on the function mode control information MODE_INFO.

[0171] For example, the function mode controller 630 can provide an enable control signal ENABLE to physical functions 1620-1 to 620-n based on the function mode control information MODE_INFO. Physical functions 1620-1 to 620-n that have received the enable control signal ENABLE can be set to active mode.

[0172] According to an embodiment, one or more physical functions 1 620-1 to physical function n 620-n that are set to active mode can perform configuration operations. During a configuration operation, information stored in the configuration spaces of physical function 610 and physical functions 1 620-1 to physical functions n 620-n can be provided to host 1000, and the configuration spaces can be set in response to a configuration request from host 1000. For example, each of the physical functions 1 620-1 to physical functions n 620-n that are set to active mode can receive a configuration request from host 1000 and can perform a configuration operation in response to the configuration request.

[0173] For example, the function mode controller 630 can provide a disable control signal DISABLE to physical functions 1620-1 to 620-n based on the function mode control information MODE_INFO. Physical functions 1620-1 to 620-n that have received the disable control signal DISABLE can be set to an inactive mode.

[0174] According to an embodiment, the function mode controller 630 can perform clock gating on one or more second physical functions 620-1 to 620-n that are set to inactive mode to reduce the power consumption of the PCIe device 600.

[0175] Therefore, according to the embodiment, by enabling or disabling physical functions 1 620-1 to physical functions n 620-n based on the function mode control information MODE_INFO received from the host 1000 using various communication technologies, these functions can be controlled according to the user's intention, and the operating expenses of the computing system including the PCIe device 600 can be reduced.

[0176] Figure 7 The physical function 700 according to an embodiment of the present disclosure is shown.

[0177] like Figure 7 The physical function 700 shown can represent Figure 6 The physical function shown is 0 610 or Figure 6 One of the physical functions 1 620-1 to 620-n shown.

[0178] Reference Figure 7 The physical function 700 may include a configuration space 710 and an operation controller 720.

[0179] Configuration space 710 may include information for configuring physical function 700. For example, configuration space 710 may include registers for configuring physical function 700 to perform operations associated with the PCIe interface. According to an embodiment, configuration space 710 may include information about the capabilities available in physical function 700.

[0180] According to an embodiment, physical function 700 can perform configuration operations in response to a configuration request from host 1000. Physical function 700 and host 1000 can communicate based on configuration space 710 set according to the configuration operations.

[0181] The operation controller 720 can control the operation of the physical function 700 based on information included in the configuration space 710. For example, the operation controller 720 can control the operation of the physical function 700 based on the configuration space 710 set according to the configuration operation settings. The operation controller 720 can also control the operation of the physical function 700 based on the field values ​​of registers set in the configuration space 710.

[0182] like Figure 6 The physical function 0 610 shown may further include configurations for managing PCIe links and for managing functions shared between physical function 0 610 and physical functions 1 620-1 to physical functions n 620-n.

[0183] Figure 8A A PCIe device in a link-up state according to an embodiment of the present disclosure is shown. Figure 8B This illustrates setting the operating mode of a physical function to active mode after the link is connected, according to an embodiment of the present disclosure. Figure 8C The present disclosure illustrates setting the operating mode of a physical function to an inactive mode under normal operating conditions, according to an embodiment of the present disclosure. Figure 8D The present disclosure illustrates setting the operating mode of a physical function to an active mode under normal operating conditions, according to an embodiment of the present disclosure.

[0184] Reference Figure 8AThe link-connected state can refer to the state of the PCIe device 600 after it is powered on or immediately following the initialization of physical functions 0 610 and 1 620-1 to 620-n. The PCIe device 600 can perform link initialization and training to enable normal packet exchange with the host 1000 via the PCIe link after the link is connected. The state of performing normal packet exchange can be defined as "L0". Link initialization and training can be performed by the Link Training and State Machine (LTSSM).

[0185] According to an embodiment, physical functions 1 620-1 to physical functions n 620-n included in the PCIe device 600 in the link-connected state can be set to an inactive mode. Therefore, physical functions 1 620-1 to physical functions n 620-n can be disabled.

[0186] refer to Figure 8B After the link is established, physical function 0 610 can receive function mode control information MODE_INFO from host 1000. Physical function 0 610 can provide function mode control information MODE_INFO to function mode controller 630.

[0187] exist Figure 8B In the example, the function mode control information MODE_INFO includes the number of physical functions to be enabled (2 in this example) and the identification information of physical functions 1 620-1 and 3 620-3 to be enabled. The function mode controller 630 can provide the enable control signal ENABLE to physical functions 1 620-1 and 3 620-3 based on the function mode control information MODE_INFO. Physical functions 1 620-1 and 3 620-3 can then set their operating mode to active mode. Therefore, physical functions 1 620-1 and 3 620-3 can be enabled.

[0188] The enabled physical function 1 620-1 and physical function 3 620-3 can then receive configuration requests from host 1000. Physical function 1 620-1 and physical function 3 620-3 can perform configuration operations in response to the configuration requests.

[0189] Reference Figure 8C After the link initialization and training are completed and physical function 0 610 is operating normally, physical function 0 610 can receive the function mode control information MODE_INFO from host 1000.

[0190] exist Figure 8CIn the example, the function mode control information MODE_INFO includes identification information for the physical function n 620-n to be enabled. The function mode controller 630 can provide an enable control signal ENABLE to the physical function n 620-n based on the function mode control information MODE_INFO. The physical function n 620-n can set its operating mode to active mode in response to the enable control signal ENABLE. Therefore, the physical function n 620-n can be enabled.

[0191] The enabled physical function n 620-n can then receive configuration requests from host 1000. Physical function n 620-n can perform configuration operations in response to configuration requests.

[0192] Reference Figure 8D After the link initialization and training are completed and physical function 0 610 is operating normally, physical function 0 610 can receive the function mode control information MODE_INFO from host 1000.

[0193] exist Figure 8D In the example, the function mode control information MODE_INFO includes identification information for the physical function 3 620-3 to be disabled. The function mode controller 630 can provide a disable control signal DISABLE to physical function 3 620-3 based on the function mode control information MODE_INFO. Physical function 3 620-3 can set its operating mode to inactive mode in response to the disable control signal DISABLE. Therefore, physical function 3 620-3 can be disabled.

[0194] The function mode controller 630 can perform clock gating on disabled physical functions 3 620-3.

[0195] Figure 9 A PCIe device 900 according to another embodiment of the present disclosure is shown.

[0196] The PCIe device 900 may further include a function status manager 910. This is because the description of physical functions 0 610, physical functions 1 620-1 to physical functions n 620-n, and the function mode controller 630 of the PCIe device 900 is consistent with... Figure 6 The corresponding features of the PCIe device 600 shown are the same, so these descriptions are omitted.

[0197] According to an embodiment, the function status manager 910 may store the status information STATUS_INFO of each of one or more second physical functions 1 620-1 to physical function n 620-n that are set to an inactive mode. For example, a disabled physical function 1 620-1 may provide the function status manager 910 and physical function 0 610 with the status information STATUS_INFO indicating an inactive state. The function status manager 910 and physical function 0 610 may store the identification information of physical function 1 620-1 and the status information STATUS_INFO of physical function 1 620-1.

[0198] According to an embodiment, physical functions 0 610 and 1 620-1 to 620-n can provide completion information in response to a REQUEST request from host 1000. The completion information may include information indicating the current status of physical functions 610 and 620-1 to 620-n. For example, the completion information may include a completion header, which includes a completion status field. The completion status field may include one of successful completion (SC), unsupported request (UR), configuration request retry status (CRS), and completer abort (CA). For example, physical functions 0 610 and 1 620-1 to 620-n can return completion information corresponding to their current status in response to a REQUEST request from host 1000.

[0199] According to an embodiment, the REQUEST request from host 1000 may include a configuration request.

[0200] In an embodiment, when PCIe device 900 receives a request REQUEST from host 1000 for a physical function that is set to inactive mode, physical function 0 610 can provide UR information to host 1000.

[0201] In another embodiment, when a request for a physical function that has been set to inactive mode is received from host 1000, the function status manager 910 may provide UR information to host 1000.

[0202] For example, host 1000 can send a request to the disabled physical function 1 620-1. In response, physical function 0 610 or function status manager 910 can provide UR information to host 1000. Host 1000, having received the UR information, can detect that physical function 1620-1 is disabled.

[0203] According to an embodiment, when host 1000 sends a request (REQUEST) to a second physical function (physical function 1 620-1 to physical function n 620-n) that is in the process of being set to an active or inactive mode, the second physical function can provide CRS information to host 1000. For example... Figure 9 This illustrates that physical function 2620-2 is set to active mode in response to the enable control signal ENABLE. Therefore, in response to receiving a REQUEST request from host 1000, physical function 2620-2 can provide CRS information to host 1000. Host 1000, having already received the CRS information, can detect that physical function 2620-2 is being configured.

[0204] In the example above, CRS information can be provided when a request is received while the physical function is being set to active mode. However, the CRS information provision operation can also be applied to situations where a request is received while the physical function is being set to inactive mode.

[0205] According to an embodiment, when one or more physical functions 1 to n (620-1 to 620-n) are set to active mode and receive a request from host 1000, the physical function can provide SC information to host 1000. For example, Figure 9 This illustrates setting physical function 3630-3 to active mode in response to the enable control signal ENABLE. When a request REQUEST is received from host 1000, physical function 3620-3 can provide SC information to host 1000. Host 1000, having received the SC information, can detect that physical function 1620-3 is enabled.

[0206] Figure 10 The process of operating a PCIe device 600 according to an embodiment of the present disclosure is illustrated.

[0207] Figure 10 The process shown can be, for example, by Figure 6 The PCIe device 600 shown Figure 9 The PCIe device 900 shown is used for execution. The following will describe... Figure 6 The PCIe 600 operation process is shown. However, the operation process can also be performed by, for example... Figure 9 The PCIe device 900 shown is used to perform this. Figure 10 In this context, the first physical function can be represented as shown in the reference above. Figure 6 or Figure 9 The aforementioned physical function 0610. In Figure 10 In this context, the second physical function can be represented as shown in the reference above. Figure 6 or Figure 9One of the physical functions 1620-1 to 620-n.

[0208] Reference Figure 10 In step S1001, the PCIe device 600 can receive function mode control information from the host 1000 through the first physical function.

[0209] The first physical function can receive function mode control information from the host 1000 using one of the following: vendor-defined messages, vendor-defined capabilities, user-defined new PCIe capabilities, Management Component Transport Protocol on System Manager Bus (SMbus) (MCTP), and MCTP on PCIe.

[0210] In step S1003, the PCIe device 600 can set the operating mode of each of the second physical functions to one of the active and inactive modes based on the functional mode control information.

[0211] Figure 11 The process of providing PCIe device completion information according to embodiments of the present disclosure is illustrated.

[0212] Figure 11 The process shown can be, for example, by Figure 9 The PCIe device 900 shown is used to perform this. Figure 11 In the middle, the first physical function can represent the above reference. Figure 6 or Figure 9 The aforementioned physical function 0610. In Figure 11 In the middle, the second physical function can represent the above reference. Figure 6 or Figure 9 One of the physical functions 1 620-1 to 620-n.

[0213] Reference Figure 11 In step S1101, the PCIe device 900 can receive a request for a second physical function from the host 1000.

[0214] In step S1103, the PCIe device 900 can determine whether the second physical function targeted by the request is set to an inactive mode.

[0215] Based on the determination result of step S1103, when the second physical function is set to inactive mode, the PCIe device 900 can provide UR information to the host 1000 in step S1105.

[0216] Based on the determination result of step S1103, when the second physical function is not set to inactive mode, the PCIe device 900 can execute step S1107.

[0217] In step S1107, the PCIe device 900 can determine whether the second physical function targeted by the request is set to active mode.

[0218] Based on the determination result of step S1107, when the second physical function is set to active mode, the PCIe device 900 can provide SC information to the host 1000 in step S1109.

[0219] Based on the determination result of step S1107, when the second physical function is not set to active mode, the PCIe device 900 can provide CRS information to the host 1000 in step S1111.

[0220] For example, when the PCIe device 900 receives a request from the host 1000 while the second physical function is being set to active mode or inactive mode, the PCIe device 900 can provide CRS information to the host 1000 in step S1111.

[0221] Figure 12 A process for controlling a physical function that is set to an active mode, according to an embodiment of the present disclosure, is illustrated.

[0222] Figure 12 The process shown can be, for example, by Figure 6 The PCIe device 600 shown Figure 9 The PCIe device 900 shown is used for execution. The following will describe... Figure 6 The operation process of the PCIe device 600 shown is illustrated. However, this operation process can also be performed by, for example... Figure 9 The PCIe device 900 shown is used to perform this. Figure 12 In the middle, the first physical function can represent the above reference. Figure 6 or Figure 9 The physical function 0 610 is described. In Figure 12 In this context, the second physical function can be represented as shown in the reference above. Figure 6 or Figure 9 One of the physical functions 1 620-1 to 620-n.

[0223] Reference Figure 12 In step S1201, the PCIe device 600 can set the operating mode of the second physical function to active mode based on the function mode control information.

[0224] In step S1203, the PCIe device 600 can receive a configuration request from the host 1000 for a second physical function that is set to active mode.

[0225] In step S1205, the PCIe device 600 can provide SC information to the host 1000.

[0226] Steps S1203 and S1205 can correspond to reference Figure 11 The steps S1101, S1107, and S1109 are described.

[0227] In step S1207, the PCIe device 600 can perform a configuration operation on the second physical function that has been set to active mode.

[0228] The PCIe device 600 can provide the host 1000 with information including a configuration space for a second physical function that is set to active mode. Additionally, the PCIe device 600 can configure the configuration space for the second physical function in response to a configuration request.

[0229] Figure 13 A process for controlling a physical function that has been set to an inactive mode, according to an embodiment of the present disclosure, is illustrated.

[0230] Figure 13 The process shown can be, for example, by Figure 6 The PCIe device 600 shown Figure 9 The PCIe device 900 shown is used for execution. The following will describe... Figure 6 The PCIe 600 operation process is shown. However, the operation process can also be performed by, for example... Figure 9 The PCIe device 900 shown is used to perform this. Figure 13 In the middle, the first physical function can represent the above reference. Figure 6 or Figure 9 The aforementioned physical function 0610. In Figure 13 In this context, the second physical function can be represented as shown in the reference above. Figure 6 or Figure 9 One of the physical functions 1 620-1 to 620-n.

[0231] Reference Figure 13 In step S1301, the PCIe device 600 can set the operating mode of the second physical function to an inactive mode based on the function mode control information.

[0232] In step S1303, the PCIe device 600 may perform clock gating on the second physical function that has been set to an inactive mode. Clock gating on the second physical function can reduce the power consumption of the PCIe device 600.

[0233] According to this disclosure, a PCIe device and its operating method can be provided that can reduce the operating cost of a computing system including a PCIe device.

[0234] Additionally, according to this disclosure, a PCIe device and its operation method for enabling or disabling PCIe functionality based on user intent can be provided.

[0235] While embodiments of this disclosure have been presented for illustrative purposes, those skilled in the art will understand that various modifications, additions, and substitutions can be made. Therefore, the invention is intended to cover all such modifications, provided they fall within the scope of the appended claims and their equivalents.

Claims

1. A high-speed peripheral component interconnect device, i.e., a PCIe device, comprising: The first physical function is to manage PCIe links and receive function mode control information from the host. Multiple secondary physical functions are assigned to multiple system images and are enabled or disabled according to their respective operating modes; as well as A function mode controller sets the operating modes of the plurality of second physical functions based on the function mode control information, wherein each of the operating modes is set to one of an active mode and an inactive mode. Specifically, one or more of the plurality of second physical functions, which are set to active mode, perform configuration operations. During these operations, configuration spaces corresponding to each of the one or more second physical functions are set, and physical resources of the PCIe device are provided to one or more system images allocated to the one or more second physical functions. Among these, the remaining second physical functions that are set to inactive mode among the plurality of second physical functions stop providing the physical resources to the remaining system images allocated to the remaining second physical functions according to clock gating.

2. The PCIe device according to claim 1, wherein the function mode control information includes at least one of the following: the number of second physical functions to be enabled among the plurality of second physical functions, the identification information of the second physical functions to be enabled, the number of second physical functions to be disabled among the plurality of second physical functions, and the identification information of the second physical functions to be disabled.

3. The PCIe device of claim 1, wherein the first physical function uses one of the following to receive the function mode control information: vendor-defined message, vendor-defined capability, user-defined new PCIe capability, Management Component Transport Protocol on System Manager Bus (SMbus) (MCTP), and MCTP on PCIe.

4. The PCIe device of claim 1, wherein when the PCIe device receives a request from the host for a second physical function of the plurality of second physical functions and the second physical function is set to an inactive mode, the first physical function provides the host with unsupported request information, i.e., UR information.

5. The PCIe device of claim 1, further comprising a function status manager that stores information relating to a second physical function among the plurality of second physical functions that is set to an inactive mode.

6. The PCIe device of claim 5, wherein when the PCIe device receives a request for the second physical function from the host and the second physical function is set to an inactive mode, the function status manager provides the host with unsupported request information, i.e., UR information.

7. The PCIe device of claim 1, wherein when one of the plurality of second physical functions receives a request from the host while the second physical function is in the process of being set to an active mode or in the process of being set to an inactive mode, the second physical function provides the host with configuration request retry status information, i.e., CRS information.

8. The PCIe device of claim 1, wherein when a second physical function of the plurality of second physical functions receives a request from the host and the second physical function is set to active mode, the second physical function provides the host with a successful completion message, i.e., an SC message.

9. The PCIe device of claim 1, wherein the configuration space includes information for configuring a corresponding second physical function; and The one or more second physical functions operate based on the information included in the configuration space.

10. The PCIe device of claim 9, wherein one or more second physical functions provide the host with information included in the configuration space.

11. A method for operating a high-speed peripheral component interconnect device, i.e., a PCIe device, the PCIe device including a first physical function and a plurality of second physical functions, the first physical function managing PCIe links connected to a host, the method comprising: The first physical function is used to receive functional mode control information from the host. Based on the functional mode control information, the operation mode of each of the plurality of second physical functions is set, and each operation mode is set as either an active mode or an inactive mode. A configuration operation is performed, in which configuration spaces corresponding to one or more second physical functions that are set to active mode are set to provide physical resources of the PCIe device to one or more system images allocated to the one or more second physical functions. Clock gating is applied to the remaining second physical functions that are set to inactive mode among the plurality of second physical functions to stop providing the physical resources to the system image allocated to the remaining second physical functions.

12. The method of claim 11, wherein the functional mode control information includes at least one of the following: the number of second physical functions to be enabled among the plurality of second physical functions, identification information of the second physical functions to be enabled, the number of second physical functions to be disabled among the plurality of second physical functions, and identification information of the second physical functions to be disabled.

13. The method of claim 11, wherein receiving the function mode control information comprises receiving the function mode control information using one of the following: vendor-defined messages, vendor-defined capabilities, user-defined new PCIe capabilities, Management Component Transport Protocol on System Manager Bus (SMbus) (MCTP), and MCTP on PCIe.

14. The method of claim 11, further comprising: In response to receiving a request from the host for a second physical function that is set to the inactive mode among the plurality of second physical functions, unsupported request information, i.e., UR information, is provided to the host.

15. The method of claim 11, further comprising: In response to receiving a request for the second physical function from the host when one of the plurality of second physical functions is being set to active or inactive mode, the system provides the host with configuration request retry status information, i.e., CRS information.

16. The method of claim 11, further comprising: In response to receiving a request from the host for a second physical function that has been set to active mode among the plurality of second physical functions, a success completion message, i.e., SC message, is provided to the host.

17. The method of claim 11, wherein the configuration operation is performed in response to receiving a configuration request for a second physical function from the host.

18. The method of claim 17, wherein performing the configuration operation further comprises: Provide the host with information included in the configuration space of the second physical function.

Citation Information

Patent Citations

  • Non-posted write transactions

    US20200004703A1

  • NVMe DIRECT VIRTUALIZATION WITH CONFIGURABLE STORAGE

    US20200042246A1

  • Multifunction device with register space for individually enabling or disabling a function of plurality of functions in response to function configuration

    US6647434B1