High-performance input-output devices supporting scalable virtualization
Techniques for scalable virtualization of an Input/Output (I/O) device are described. An electronic device composes a virtual device comprising one or more assignable interface (AI) instances of a plurality of AI instances of a hosting function exposed by the I/O device. The electronic device emulates device resources of the I/O device via the virtual device. The electronic device intercepts a request from the guest pertaining to the virtual device, and determines whether the request from the guest is a fast-path operation to be passed directly to one of the one or more AI instances of the I/O device or a slow-path operation that is to be at least partially serviced via software executed by the electronic device. For a slow-path operation, the electronic device services the request at least partially via the software executed by the electronic device.
1. An apparatus comprising:
a plurality of backend resources; and
one or more physical function base address registers (PF-BARs) to store one or more base addresses of one or more address ranges to be mapped to a plurality of memory-mapped input/output (MMIO) registers, including a first MMIO register to be accessed for a direct-path operation and a second MMIO register to be accessed for an intercepted-path operation; wherein hardware is to provide guest physical address to host physical address translation for the first MMIO register but not for the second MMIO register;
access for the direct-path operation is to be mapped by a virtual machine monitor (VMM) to an interface for a virtual device, the interface to be composed of one or more of the plurality of backend resources and to be identified by a process address-space identifier (PASID); and
access for the intercepted-path operation is to be intercepted by the VMM for emulation.
2. The apparatus of claim 1 , wherein the one or more PF-BARs includes a plurality of variable size PF-BARs.
3. The apparatus of claim 1 , wherein the one or more PF-BARs includes a plurality of non-contiguous PF-BARs.
4. The apparatus of claim 1 , wherein the one or more PF-BARs include at least one Peripheral Component Interconnect Express (PCIe) base address register.
5. The apparatus of claim 1 , wherein the access for the intercepted-path operation is to be intercepted by the VMM for emulation of a configuration space of the virtual device.
6. The apparatus of claim 1 , wherein the plurality of backend resources includes at least one of a transmission/reception (Tx/Rx) queue, a command queue, a Field Programmable Gate Array (FPGA) context, a set of one or more processing units, a Graphics Processing Unit (GPU) context, and a general-purpose computing on graphics processing unit (GPGPU) context.
7. A method comprising:
storing, in one or more physical function base address registers (PF-BARs), one or more base addresses of one or more address ranges to be mapped to a plurality of memory-mapped input/output (MMIO) registers, including a first MMIO register to be accessed for a direct-path operation and a second MMIO register to be accessed for an intercepted-path operation;
performing, by hardware, guest physical address to host physical address translation for the first MMIO register but not for the second MMIO register;
mapping, by a virtual machine monitor (VMM), access for the direct-path operation to an interface for a virtual device, the interface to be composed of one or more of a plurality of backend resources and to be identified by a process address-space identifier (PASID); and
intercepting, by the VMM, access for the intercepted-path operation for emulation.
8. The method of claim 7 , wherein the one or more PF-BARs includes a plurality of variable size PF-BARs.
9. The method of claim 7 , wherein the one or more PF-BARs includes a plurality of non-contiguous PF-BARs.
10. The method of claim 7 , wherein the one or more PF-BARs include at least one Peripheral Component Interconnect Express (PCIe) base address register.
11. The method of claim 7 , wherein the access for the intercepted-path operation is to be intercepted by the VMM for emulation of a configuration space of the virtual device.
12. The method of claim 7 , wherein the plurality of backend resources includes at least one of a transmission/reception (Tx/Rx) queue, a command queue, a Field Programmable Gate Array (FPGA) context, a set of one or more processing units, a Graphics Processing Unit (GPU) context, and a general-purpose computing on graphics processing unit (GPGPU) context.
13. A non-transitory machine-readable medium storing instructions, which when executed by a machine, cause the machine to perform a method comprising:
storing, in one or more physical function base address registers (PF-BARs), one or more base addresses of one or more address ranges to be mapped to a plurality of memory-mapped input/output (MMIO) registers, including a first MMIO register to be accessed for a direct-path operation and a second MMIO register to be accessed for an intercepted-path operation;
performing, by hardware, guest physical address to host physical address translation for the first MMIO register but not for the second MMIO register;
mapping, by a virtual machine monitor (VMM), access for the direct-path operation to an interface for a virtual device, the interface to be composed of one or more of a plurality of backend resources and to be identified by a process address-space identifier (PASID); and
intercepting, by the VMM, access for the intercepted-path operation for emulation.
14. The non-transitory machine-readable medium of claim 13 , wherein the one or more PF-BARs includes a plurality of variable size PF-BARs.
15. The non-transitory machine-readable medium of claim 13 , wherein the one or more PF-BARs includes a plurality of non-contiguous PF-BARs.
16. The non-transitory machine-readable medium of claim 13 , wherein the one or more PF-BARs include at least one Peripheral Component Interconnect Express (PCIe) base address register.
17. The non-transitory machine-readable medium of claim 13 , wherein the access for the intercepted-path operation is to be intercepted by the VMM for emulation of a configuration space of the virtual device.
18. The non-transitory machine-readable medium of claim 13 , wherein the plurality of backend resources includes at least one of a transmission/reception (Tx/Rx) queue, a command queue, a Field Programmable Gate Array (FPGA) context, a set of one or more processing units, a Graphics Processing Unit (GPU) context, and a general-purpose computing on graphics processing unit (GPGPU) context.