Techniques for supporting for demand paging

• US 10,114,758 B2
• Filed: 09/13/2013
• Issued: 10/30/2018
• Est. Priority Date: 09/13/2013
• Status: Active Grant

• Current Assignee: NVIDIA Corporation
• Original Assignee: NVIDIA Corporation
• Inventors: Duncan, Samuel H., Duluk, Jr., Jerome F., Evans, Jonathon Stuart Ramsay, Deming, James Leroy
• Primary Examiner(s): Birkhimer, Christopher D
• Application Number:
  US14/026,806
• Publication Number:
  US 20150082001A1
• Time in Patent Office: 1,873 Days
• Field of Search: None
• CPC Class Codes: G06F 12/08 in hierarchically structured memory systems, e.g. virtual memory systems
• CPC Class Codes: G06F 12/1009 using page tables, e.g. page table structures

Abstract

One embodiment of the present invention includes techniques to support demand paging across a processing unit. Before a host unit transmits a command to an engine that does not tolerate page faults, the host unit ensures that the virtual memory addresses associated with the command are appropriately mapped to physical memory addresses. In particular, if the virtual memory addresses are not appropriately mapped, then the processing unit performs actions to map the virtual memory address to appropriate locations in physical memory. Further, the processing unit ensures that the access permissions required for successful execution of the command are established. Because the virtual memory address mappings associated with the command are valid when the engine receives the command, the engine does not encounter page faults upon executing the command. Consequently, in contrast to prior-art techniques, the engine supports demand paging regardless of whether the engine is involved in remedying page faults.

Claims

1. A computer-implemented method for supporting demand paging, the method comprising:

receiving one or more requests for a copy engine within a processing subsystem to perform one or more operations, wherein the one or more requests comprise one or more virtual memory addresses that correspond to the one or more operations;

prior to transmitting any request included in the one or more requests to the copy engine, ensuring that the processing subsystem includes a memory mapping for each of the one or more virtual memory addresses; and

transmitting the one or more requests to the copy engine for processing.

2. The method of claim 1, wherein ensuring that the processing subsystem includes a memory mapping for one or more virtual memory addresses comprises ensuring that the processing subsystem includes a memory mapping for a first virtual memory address by transmitting a first probe associated with the first virtual memory address to a memory management unit, and receiving a response to the first probe.

3. The method of claim 2, wherein transmitting the first probe comprises updating a run-list of processes scheduled to run on the processing subsystem to include a command to transmit the first probe to the memory management unit, wherein the run-list reflects one or more processes that perform at least one update to the memory mapping.

4. The method of claim 3, wherein ensuring that the processing subsystem includes the memory mapping for the first virtual memory further comprises determining that an update to the memory mapping is scheduled to occur subsequent to the transmission of the first probe, retransmitting the first probe to the memory management unit, and receiving a second response to the first probe.

5. The method of claim 2, wherein receiving a response to the first probe comprises receiving a first acknowledgement indicating that the processing subsystem includes the memory mapping for the first virtual memory address.

6. The method of claim 5, wherein ensuring that the processing subsystem includes a memory mapping for a second virtual memory address included in the one or more virtual memory addresses comprises:

transmitting a second probe associated with the second virtual memory address to the memory management unit;

determining that a page fault associated with the second probe has been resolved;

retransmitting the first probe to the memory management unit;

receiving a second acknowledgement indicating that the processing subsystem includes the memory mapping for the first virtual memory address;

retransmitting the second probe to the memory management unit; and

receiving a third acknowledgement indicating that the processing subsystem includes the memory mapping for the second virtual memory address.

7. The method of claim 6, wherein the memory management unit generates the page fault based on the second probe.

8. The method of claim 1, wherein ensuring that the processing subsystem includes a memory mapping for a first virtual address included in the one or more virtual memory addresses comprises:

transmitting a first probe associated with the first virtual memory address to a memory management unit;

determining that a page fault associated with the first probe has been resolved;

retransmitting the first probe to the memory management unit; and

receiving an acknowledgement indicating that the processing subsystem includes the memory mapping for the first virtual memory address.

9. The method of claim 8, wherein the memory management unit generates the page fault based on the first probe.

10. The method of claim 9, wherein determining that the page fault is resolved comprises receiving an interrupt.

11. The method of claim 1, wherein the operation comprises a write operation to a first virtual memory address included in the one or more virtual memory addresses, and ensuring that the processing subsystem includes a memory mapping for the first virtual memory address comprises:

transmitting a probe associated with the write operation to a memory management unit; and

receiving a response to the probe indicating that the processing subsystem includes the memory mapping for the first virtual address and that the memory mapping for the first virtual address includes an activated write enable bit.

23. The method of claim 1, wherein the copy engine comprises an engine unable to service a page fault caused by an absence of the memory mapping in the processing subsystem.

24. The method of claim 1, further comprising ensuring that no request is transmitted to the copy engine prior to ensuring that the processing subsystem includes a memory mapping for each virtual memory address associated with the one or more requests.

26. The method of claim 1, wherein ensuring that the processing subsystem includes a memory mapping for each of the one or more virtual memory addresses is performed via a host coupled to the copy engine.

12. A subsystem to support demand paging, the subsystem comprising:

a memory that includes a page table;

a memory management unit coupled to the memory, wherein the memory management unit;

processes requests for translating virtual memory addresses to physical memory addresses, and

manages the page table;

a copy engine coupled to the memory management unit, wherein the copy engine performs operations associated with virtual memory addresses; and

a host coupled to the copy engine and to the memory management unit, wherein the host;

receives one or more requests for the copy engine to perform one or more operations, wherein the one or more requests comprise one or more virtual memory addresses that correspond to the one or more operations;

prior to transmitting any request included in the one or more requests to the copy engine, ensures that the page table includes a memory mapping for each of the one or more virtual memory addresses; and

transmits the one or more requests to the copy engine for processing.

13. The subsystem of claim 12, wherein ensuring that the page table includes a memory mapping for a first virtual memory address included in the one or more virtual memory addresses comprises transmitting a first probe associated with the first virtual memory address to the memory management unit, and receiving a response to the first probe.

14. The subsystem of claim 13, wherein receiving a response to the first probe comprises receiving a first acknowledgement indicating that the page table includes the memory mapping for the first virtual memory address.

15. The subsystem of claim 14, wherein ensuring that the page table includes a memory mapping for a second virtual memory address included in the one or more virtual memory addresses comprises:

transmitting a second probe associated with the second virtual memory address to the memory management unit;

determining that a page fault associated with the second probe has been resolved;

retransmitting the first probe to the memory management unit;

receiving a second acknowledgement indicating that the page table includes the memory mapping for the first virtual memory address;

retransmitting the second probe to the memory management unit; and

receiving a third acknowledgement indicating that the page table includes the memory mapping for the second virtual memory address.

16. The subsystem of claim 15, wherein the memory management unit generates the page fault based on the second probe.

17. The subsystem of claim 12, wherein ensuring that the page table includes a memory mapping for a first virtual address included in the one or more virtual memory addresses comprises:

transmitting a first probe associated with the first virtual memory address to the memory management unit;

determining that a page fault associated with the first probe has been resolved;

retransmitting the first probe to the memory management unit; and

receiving an acknowledgement indicating that the page table includes the memory mapping for the first virtual memory address.

18. The subsystem of claim 17, wherein the memory management unit generates the page fault based on the first probe.

19. The subsystem of claim 18, wherein determining that the page fault is resolved comprises receiving an interrupt.

20. The subsystem of claim 12, wherein the operation comprises a write operation to a first virtual memory address included in the one or more virtual memory addresses, and ensuring that the page table includes a memory mapping for the first virtual memory address comprises:

transmitting a probe associated with the write operation to the memory management unit; and

receiving a response to the probe indicating that the page table includes the memory mapping for the first virtual address and that the memory mapping for the first virtual address includes an activated write enable bit.

25. The subsystem of claim 12, wherein ensuring that the page table includes a memory mapping for each of the one or more virtual memory addresses comprises issuing a probe associated with the one or more virtual memory addresses to the memory management unit to test whether the memory management unit supports a translation for the one or more virtual memory addresses.

21. A computing device, comprising:

a memory that includes a page table;

a memory management unit that;

processes requests for translating virtual memory addresses to physical memory addresses, and

manages the page table;

a copy engine that performs operations associated with virtual memory addresses; and

a host that;

receives one or more requests for the copy engine to perform one or more operations, wherein the one or more requests associated with comprise one or more virtual memory addresses that correspond to the one or more operations;

prior to transmitting any request included in the one or more requests to the copy engine, ensures that the page table includes a memory mapping for each of the one or more virtual memory addresses; and

transmits the one or more requests to the copy engine for processing.

22. The computing device of claim 21, wherein ensuring that the page table includes a memory mapping for a first virtual address included in the one or more virtual memory addresses comprises:

transmitting a first probe associated with the first virtual memory address to the memory management unit;

determining that a page fault associated with the first probe has been resolved;

retransmitting the first probe to the memory management unit; and

receiving an acknowledgement indicating that the page table includes the memory mapping for the first virtual memory address.