Kernel software RAID support for direct-access file systems

1. A method, comprising:

context switching, by a processing device, into a kernel associated with a persistent memory to provide a write confirmation to a client device requesting data be written;

updating, in response to the context switching, a parity block associated with a page of the persistent memory; and

providing, in response to updating the parity block, the write confirmation to the client device.

2. The method of claim 1 , further comprising detecting a write fault associated with the page of the persistent memory.

3. The method of claim 2 , further comprising:

resolving the write fault; and

writing, in response to resolving the write fault, the data to the page of the persistent memory.

4. The method of claim 2 , further comprising:

generating, in response to detecting the write fault, an undo log corresponding to current data of the page of the persistent memory; and

copying the undo log to a second persistent memory.

5. The method of claim 4 , further comprising:

restoring the current data from the undo log; and

providing an indication to the client device that the data was not written.

6. The method of claim 2 , further comprising:

receiving a request to write the data from the client device; and

generating the write fault in response to attempting to write the data.

7. The method of claim 2 , wherein detecting the write fault associated with the page of the persistent memory comprises determining that the page was memory mapped and is read-only.

8. A system, comprising:

a persistent memory to store data; and

a processing device, operatively coupled to the persistent memory, the processing device to:

context switch into a kernel associated with the persistent memory to provide a write confirmation to a client device requesting data be written;

update, in response to the context switching, a parity block associated with a page of the persistent memory; and

provide, in response to the update of the parity block, the write confirmation to the client device.

9. The system of claim 8 , wherein the processing device is further to resolve a write fault, and wherein to resolve the write fault, the processing device is to:

generate, in response to detecting the write fault, an undo log corresponding to current data of the page; and

store the undo log in a second persistent memory.

10. The system of claim 9 , wherein the processing device is further to copy the undo log to another persistent memory.

11. The system of claim 10 , wherein a redundant array of independent disk (RAID) leg storing the page fails or suffers a power failure after writing the data and before writing the parity block, and on recovery from failing, and wherein the processing device is further to:

restore the current data from the undo log; and

provide an indication to the client device that the data was not written.

12. The system of claim 8 , wherein an application or operating system of the client device fails after writing the data and before writing the parity block, and on recovery from failing, and wherein the processing device is further to:

provide control to the kernel; and

update the parity block.

13. The system of claim 2 , wherein the processing device is further to:

receive a request to write the data from the client device; and

generate the write fault in response to attempting to write the data.

14. The system of claim 9 , wherein to detect the write fault associated with the page, and wherein the processing device is further to determine that the page was memory mapped and is read-only.

15. A method, comprising:

providing an interposition driver; and

context switching, by a processing device, into a kernel associated with a persistent memory using the interposition driver to create a consistent view of the persistent memory.

16. The method of claim 15 , further comprising:

providing a write confirmation to a client device requesting data be written; and

updating, in response to the context switching, a parity block associated with a page of the persistent memory.

17. The method of claim 16 , wherein a redundant array of independent disk (RAID) leg storing the page fails or suffers a power failure and the method further comprises:

restoring current data from an undo log; and

provide an indication to the client device that the data was not written.

18. The method of claim 16 , wherein an application or operating system of the client device fails after writing the data and before writing the parity block, and on recovery from failing, the method further comprising:

providing control to the kernel; and

updating the parity block.